ArticlePDF Available

Data, Information, and Knowledge: Have We Got It Right?

Authors:

Abstract and Figures

Economists make the unarticulated assumption that information is something that stands apart from and is independent of the processor of information and its internal characteristics. We argue that they need to revisit the distinctions they have drawn between data, information, and knowledge. Some associate information with data, and others associate information with knowledge. But since none of them readily conflates data with knowledge, this suggests too loose a conceptualisation of the term ‘information’. We argue that the difference between data, information, and knowledge is in fact crucial. Information theory and the physics of information provide us with useful insights with which to build an economics of information appropriate to the needs of the emerging information economy. Copyright Springer-Verlag Berlin/Heidelberg 2004
Content may be subject to copyright.
Data, information and knowledge: have we got it
right?
Max Boisot (boisot@attglobal.net)
Researcher (IN3-UOC)
Sol Snider Center for Entrepreneurial Research (The Wharton School, University of Pennsylvania)
Agustí Canals (acanalsp@uoc.edu)
Researcher (IN3-UOC)
Director of Information and Communications Sciences Studies (UOC)
Working Paper Series WP04-002
Submission date: November 2003
Published in: February 2004
Internet Interdisciplinary Institute (IN3): http://www.uoc.edu/in3/eng/index.htm
ABSTRACT
Economists make the unarticulated assumption that information is something that stands apart
from and is independent of processors of information and their inherent characteristics. We
argue that they need to revisit the distinctions they have drawn between data, information and
knowledge. While some associate information with data, others associate it with knowledge. But
since few readily associate data with knowledge, this suggests too loose a conceptualisation of
the term 'information'. We argue that the difference between data, information and knowledge is
in fact crucial. Information theory and the physics of information provide us with useful insights
with which to build an economics of information appropriate to the needs of the emerging
information economy.
KEYWORDS
Information, Knowledge, Economics of Information, Information Theory, Physics of Information
SUMMARY
1. Introduction
2. Conceptualizing the Issue
3. Information, Individuals, and Organizations
4. The Contribution of Information Theory
5. The Physics of Information
6. An economic interpretation of the principle of least action
7. Implications
8. Conclusion
References
To cite this document, you could use the following reference:
BOISOT, Max; CANALS, Agustí (2004). Data, information and knowledge: have we got it right? [Online Working
Paper. Submission date: 17th November, 2003]. IN3: UOC. (Working Paper Series; DP04-002) [Date of citation:
dd/mm/yy]
<http://www.uoc.edu/in3/dt/20388/index.html>
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
2004 by Max Boisot and Agustí Canals
2004 by FUOC -2-
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
Data, information and knowledge: have we got it
right?
MAX BOISOT
IN3, Universitat Oberta de Catalunya, Barcelona, Catalonia, Spain.
Sol Snider Center for Entrepreneurial Research, The Wharton School, University of
Pennsylvania.
e-mail: boisot@attglobal.net
AGUSTÍ CANALS1
IN3, Universitat Oberta de Catalunya, Barcelona, Catalonia, Spain.
ESADE Business School, Universitat Ramon Llull, Barcelona, Catalonia, Spain.
e-mail: acanalsp@uoc.edu
Version 3, 17th November, 2003
Abstract
Economists make the unarticulated assumption that information is something that stands
apart from and is independent of the processor of information and its internal
characteristics. We argue that they need to revisit the distinctions they have drawn between
data, information, and knowledge. Some associate information with data, and others
associate information with knowledge. But since none of them readily conflates data with
knowledge, this suggests too loose a conceptualisation of the term ‘information’. We argue
that the difference between data, information, and knowledge is in fact crucial. Information
theory and the physics of information provide us with useful insights with which to build an
economics of information appropriate to the needs of the emerging information economy.
1 To whom proofs should be addressed. Present address: Av. Tibidabo, 39-43. 08035 Barcelona, Catalonia,
Spain. E-mail: acanalsp@uoc.edu
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
JEL Classification: A12, D20, D80, M21
Key words: Information—Knowledge—Economics of Information—Information
Theory—Physics of Information
1. Introduction
Effective cryptography protects information as it flows around the world. Encryption, by
developing algorithms that bury information deeply in data, provides “the lock and keys” of
the information age (Singh, 1999, p. 293). Thus while the data itself can be made “public”
and hence freely available, only those in possession of the “key” are in a position to extract
information from it (Singh, 1999). Cryptography, in effect, exploits the deep differences
between data and information.
Knowledge and information are not the same thing, either. Imagine, for example, receiving
an encrypted message for which you possess the key and from which you extract the
following information: “The cat is tired”. Unless you possess enough contextual
background knowledge to realize that the message refers to something more than an
exhausted cat – possibly a Mafia boss, for example - you may not be in a position to react in
an adaptive way. To Understand the sentence is not necessarily to understand the message.
Only prior knowledge will allow a contextual understanding of the message itself, and the
message, in turn will carry information that will modify that knowledge. Clearly, then,
information and knowledge must also be distinguished from one another.
In everyday discourse, the distinction between data and information, on the one hand, and
between information and knowledge, on the other, remains typically vague. At any given
moment, the terms data and information will be used interchangeably; whereas at another,
information will be conflated with knowledge. Although few people will argue that
knowledge can ever be reduced to data, the two terms are unwittingly brought into a forced
marriage by having the term information act as an informal go-between. The growing
commercial interest in cryptography, however, suggests innumerable practical
circumstances in which the need to distinguish between the three terms is becoming
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
compelling. But if the distinction works in practice, does it work in theory? This is the
question that our paper addresses.
Beginning with the second half of the twentieth century, a number of economists –
Koopmans, Marschak, Hurwicz, and Arrow – began to concern themselves with the nature
of the economic agent as a ”rational information processor”. Since that time, information
has become acknowledged as the key generator of wealth in post-industrial societies. We
might therefore reasonably assume that, over the past fifty years, mainstream economists,
concerned as they are with wealth creation, would have developed a conceptual approach to
information that reflected its growing importance to their field.
In this paper, we shall argue that they have some way to go. Both Stiglitz and Lamberton
have noted how, even at the end of the twentieth century, the economic profession’s
conviction that there can be an ‘economics of information’ still has to reckon with the lack
of any consensus as to what specifically it should cover (Stiglitz, 2000; Lamberton, 1998).
As Arrow has commented, “ It has proved difficult to frame a general theory of information
as an economic commodity, because different kinds of information have no common unit
that has yet been identified” (Arrow, 1973, p. iii). In fact, Arrow believed that such units
were undefinable (Arrow, 1996)2.Economics, then, is still looking for a way of thinking
about information that is adapted both to its own analytical needs as well as to the needs of
the emerging information economy.
For this reason, we can support Lamberton’s plea that we abandon a unitary and all-purpose
concept of information and develop instead a taxonomy based on significant characteristics
of information (Lamberton, 1996, pp. xx-xxii). However, descriptions will not, by
themselves, build viable taxonomies. Only adequate theorizing will tell us what
characteristics will be taxonomically significant. Here we initiate some necessary theorizing
that takes as its focus the differences between data, information and knowledge. We shall
proceed as follows. First, in the next section (2) we develop a simple conceptual scheme to
inform our subsequent discussion. In section 3, we briefly look at how the economic and
organizational sciences have dealt with these differences. Both have tended to conflate
information and knowledge and to ignore the role of data. In section 4, we examine what
information theory adds to the picture. In section 5, we broaden our analysis by introducing
2 James Boyle has analyzed the incoherence of information economics over a period of fifty years in his
Shamans, Software and Spleens (1996).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
concepts from a new field, the physics of information. Here, the conflation has been of
information with data rather than with knowledge – the ”observer” in physics need have no
cognitive capacities as such, only a perceptual ability to distinguish between simple
physical states. In section 6, with the help of a simple diagram somewhat reminiscent of a
production function, we briefly illustrate how the distinction between data, information, and
knowledge might be exploited in economic theorizing. In section 7, we explore the
implications of our comparative analysis for an economics of information and put forward
three propositions. A conclusion follows in section 8.
2. Conceptualizing the Issue
Consider the way in which economists theorize about information in game theory. Game
theory deals with a situation in which knowledge is either taken as being asymmetrically
distributed or is taken to be common knowledge (Aumann, 1976; Hargreaves Heap and
Varoufakis, 1995), the Nash concept specifying both the game’s information requirements
and the conditions of its transmission (Kuhn, 1962; Myerson, 1999). These were hardly
models of realism. Yet as game theory evolved in the 1980s and 1990s, it imposed ever-less
plausible cognitive conditions on economic agents (Binmore, 1990), reflecting its allegiance
to neoclassical concepts of information, knowledge, and computability, as well as to the
Arrow-Debreu model of Walrasian general equilibrium (Mirowski, 2002).
How, for example, does game theory deal with the situation in which repeated games unfold
under dynamic conditions of information diffusion? Here, information is asymmetrically
distributed when the first game takes place and is common knowledge by the time the last
game occurs. This situation can also be made to work in reverse. Information can start off
as common knowledge in a first game and become asymmetrically distributed by the time
the last one occurs. Williamson takes this latter outcome as resulting from a "fundamental
transformation" wherein an initial large-numbers bargaining process by degrees gets
transformed into a small-numbers bargaining process. Here, contract renegotiation involves
an ever-decreasing number of players on account of asymmetrically distributed learning
opportunities combined with the effects of information impactedness (Williamson, 1985).
This second situation might then count as an instance of repeated games in which
information gets differentially "impacted" (Williamson’s term) among the different players
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
according to their respective learning abilities as the games unfold to give them anything
but “common knowledge”.
How should data, information and knowledge be conceptualized to account for this?
Economists struggle. Or not: Hirschleifer and Riley, for example, in their widely read and
popular text, The Analytics of Uncertainty and Information (1992), hardly deal with
definitional issues at all. Taking information to be an input into decision-making, the
authors identify the lack of objective information and the lack of information about the
future as the key problems they wish to address. A third problem, the limitation of human
understanding when dealing with information, the authors choose to ignore on the ground
that their intention is to “model economics, not psychology” (Hirshleifer and Riley, 1992, p.
8). Clearly here, the unarticulated assumption – implicitly endorsed by Shannon’s
information theory (Mirowski, 2002) - is that information is something that stands apart
from and is independent of the processor of information and its internal characteristics.
Information itself is loosely defined as either “knowledge”—ie, as a “stock” - or as an
“increment to the stock of knowledge”—ie, as “news” or “message”. Like information,
knowledge and/or news are assumed to exist independently of a knower or a receiver of
news. The tacit assumption that information and knowledge are "things" is widely held. It
is, however, a strong assumption, and therefore one that could only follow from an
appropriate conceptualization of information, of knowledge, and of the ways in which they
relate to each other. Yet nowhere in Hirschleifer and Riley's book is it possible to find a
treatment of information and knowledge that is rigorous enough to serve as a basis for such
an assumption and for the economic analysis that builds on it.
If Hirschleifer and Riley associate information with knowledge, two other economists,
Shapiro and Varian, taking information to be anything that can be digitized, associate it
with data (Shapiro and Varian, 1999). Since data is ”thing-like”, it follows that information
is also ”thing-like”, a shared property that allows these authors to claim that the new
information economy can draw on the same economic laws as those that govern the energy
economy. Here again, the way that data and information relate to one another is ignored.
Yet, although data might be taken as thing-like and given - that is after all what the roots of
the term datum (what is given) imply - what is taken to constitute information is always
evolving to reflect the changing relationship between agents and data. Thus, whereas the
analysis of data lends itself to the application of comparative statics and can be linearized,
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
the analysis of information requires the examination of complex feedback loops and the
application of nonlinear dynamics. The view that information is itself a thing rather than a
relation points to the survival of essentialist thinking in economics, and of a concern with
being rather than with becoming (Prigogine, 1980).
Since the distinction between data, information, and knowledge is the focus of this paper,
we now briefly discuss how it might be approached.
Data can be treated as originating in discernible differences in physical states-of-the-world
– that is, states describable in terms of space, time, and energy. Anything that can be
discerned at all is discerned on the basis of such differences (Rosen, 1991) and is discerned
by agents (Derrida, 1967; Deleuze, 1969). Agents are bombarded by stimuli from the
physical world, not all of which are discernable by them and hence not all of which register
as data for them. Much neural processing has to take place between the reception of a
stimulus and its sensing as data by an agent (Kuhn, 1974). It takes energy for a stimulus to
register as data, the amount of energy being a function of the sensitivity of the agent’s
sensory apparatus (Crary, 1999). Information constitutes those significant regularities
residing in the data that agents attempt to extract from it. It is their inability to do so
costlessly and reliably that gives encryption its power and that makes the distinction
between data and information meaningful. For if data and information were the same thing,
the effective encryption of messages – ie, the concealing information in data in such a way
that third parties cannot extract it – would be impossible (Singh, 1999).
What constitutes a significant regularity, however, can only be established with respect to
the individual dispositions of the receiving agent. Information, in effect, sets up a relation
between in-coming data and a given agent. Only when what constitutes a significant
regularity is established by convention, can information appear to be objective – and even
then, only within the community regulated by the convention. Finally, knowledge is a set of
expectations held by agents and modified by the arrival of information (Arrow, 1984).
These expectations embody the prior situated interactions between agents and the world - in
short, the agent's prior learning. Such learning need not be limited – as required by the
theory of rational expectations (Muth, 1961) – to models specifically relevant to the
expectations to which they give rise.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
To summarize, we might say that information is an extraction from data that, by modifying
the relevant probability distributions, has a capacity to perform useful work on an agent's
knowledge base. The essential relationships between data, information and knowledge are
depicted in Figure 1. The diagram indicates that agents operate two kinds of filters in
converting incoming stimuli into information. Perceptual filters first orient the senses to
certain types of stimuli that operate within a given physical range. Only stimuli passing
through this initial filter get registered as data3. Conceptual filters then extract information-
bearing data from what has been so registered. Both types of filters get ”tuned” by the
agents’ cognitive and affective expectations (Clark, 1997; Damasio, 1999), shaped as these
are by prior knowledge, to act selectively on both stimuli and data.
Fig. 1. The Agent-in-the-World
The schema depicted in figure 1 allows us to view data, information and knowledge as
distinct kinds of economic goods, each possessing a specific type of utility. The utility of
data resides in the fact that it can carry information about the physical world; that of
information, in the fact that it can modify an expectation or a state of knowledge; finally,
3 Roland Omnes, the philosopher of quantum mechanics, understands data thus: “In order to understand what
a measurement is, it would be helpful first to make a distinction between two notions that are frequently
confused: an experiment’s (concrete) data and its (meaningful) result. The data are for us a macroscopic
classical fact: thus when we see the numeral 1 on the Geiger counter’s screen, this is the datum. The result is
something different, for it is a strictly quantum property, almost invariably pertaining only to the microscopic
world, meaning that a radioactive nucleus disintegrated, for example, or providing a component of a particle’s
spin. The datum is a classical property concerning only the instrument; it is the expression of a fact. The result
concerns a property of the quantum world. The datum is an essential intermediary for reaching a result.”
(Omnes, 1999, author's italics).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
that of knowledge in the fact that it allows an agent to act in adaptive ways in and upon the
physical world. Telephone books are paradigmatically data goods; specialized newsletters,
being more selective, exemplify information goods; and brain surgery can be thought of as
a knowledge good. We shall not elaborate further on these different types of good.
3. Information, Individuals, and Organizations
Most of what modern economics has to say about knowledge and information originates in
the tradition of methodological individualism (Hodgson, 1988; 1993)4. This tradition takes
the individual human being, Homo Economicus, as the fundamental unit of analysis. The
origins of methodological individualism are deeply rooted in the Anglo-Saxon political
economy tradition that goes back to Hobbes and to Locke (MacPherson, 1962). The central
challenge was to protect the rationality postulate inherited from the Enlightenment from the
centrifugal tendencies at work when varied and complex individuals pursue their own
interests in both markets and organizations. The socialist calculation controversy of the
1930s opposed those who believed that rationality was best preserved through a central
planning mechanism – a metamorphosis of Walras’s auctioneer – to those who believed in
preserving it through a decentralized market mechanism. The concern with the
computational efficiency of either mechanism placed the focus on the coordinating role of
”knowledge” and ”information” (Von Mises, 1969; Lange and Taylor, 1938; Hayek, 1999)
and on the computational characteristics of different types of economic agency – the state,
the firm, the individual. More recent attempts to deal with these threats to economic
rationality have resulted in a kind of methodological “cyborgism” (Mirowski, 2002) that
builds information structures both above and between agents5.
What are the computational requirements of the neoclassical rationality postulate? Most
relevant from our perspective is the fact that, whatever the type of economic agent involved,
it is not subject to communicative or data processing limitations. The information
environment in which it operates is free of noise and friction – well-structured information
is instantaneously available in the form of prices and these fully capture the relevant
4 The Marxist tradition in economics has an even less tenable position on information than does neoclassical
economics. In the Marxist tradition, information asymmetries are deliberately created for the purposes of
exploitation. Information goods are or should be, by their nature, free goods. See Marx (1867).
5 We are indebted to a reviewer of this paper for this observation.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
attributes of trades into which the actor enters6. What does an individual economic agent
actually do with information? He computes in order to take decisions (Hurwicz, 1969;
Marschak, 1974; Arrow, 1973). His computational abilities are unbounded and it enjoys
both infinite memory and infinite foresight (Stiglitz, 1983). It follows, therefore, that such
an agent does not need to learn much (Hodgson, 1999). It is the frictionless ease with which
the rational economic agent is able to compute and communicate that qualifies him as
“Newtonian”.
It is by now well established that Homo Economicus has not served neoclassical economics
well as a model of the way in which real human beings process and transmit data. These
agents are bounded in their rationality (Miller, 1956; Simon, 1945) and are subject to
systematic cognitive biases (Kahneman and Tversky, 1982; Kahneman, 1994; Bruner,
1974; Jung, 1971). It was a coming to terms with the cognitive limitations of individual and
group processes that gradually turned economics into what Mirowski calls a “Cyborg
science”, with Hayek as its prophet (Mirowski, 2002; Hayek, 1999; Gottinger, 1983;
Makowski and Ostroy, 1993; Smith, 1991; Lavoie, 1985; Weimer and Palermo, 1982).
Evolutionary economics developed a more realistic – not to say “naturalistic” (Quine, 1969)
- perspective on the role of knowledge in human affairs than has orthodox economics
(Hodgson, 1993; Vromen, 1995; Hamilton, 1991; Nelson, 1994). The omniscience of agents
is out! For Nelson and Winter (1982), for example, the routinization of firm activities is a
response to information complexity. It is in rules and routines that a firm’s knowledge is
deemed to be stored. These then become the units of selection in an evolutionary process.
Yet, as Fransman points out, the tight coupling of information and knowledge that is
implied – with knowledge becoming little more than processed information - is unrealistic,
since different agents may extract different knowledge from the same information
(Fransman, 1998). Indeed, the variety on which evolutionary selection is effectively
predicated, depends on it! Fransman himself goes on to associate information with data – a
tight coupling in the other direction - and knowledge with belief.
If economists of different stripes have tended to conflate knowledge and information,
sociologists, by contrast, have been more concerned with knowledge alone. Furthermore,
sociology's point of departure is not the asocial atomized individual, but the embedded
6 As Koopmans put it “The economies of information handling are secured free of charge (Koopmans, 1957).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
socialized actor (Granovetter, 1985). Mead, for example, emphasized “the temporal and
logical pre-existence of the social process to the self-conscious individual that arises in it”
(Mead, 1934, p. 186). Thus, in contrast with the methodological individualism of
economics, sociology “problematizes” the individual, often adopting a Vygotskian view
that society should be the point of departure for looking at the evolution of human
information processing capacities (Vygotsky, 1986). Durkheim and Mauss (1903), for
example, analyzed primitive classification schemes as collective forms of representation.
Sociology, then, typically starts with a multiple-actor perspective and gradually homes in on
the single actor.
Finally, the sociology of knowledge tradition emphasizes the way in which power shapes
collective representations (Mannheim, 1960; Habermas, 1987). By viewing human
rationality as socially bounded by power and institutions (DiMaggio and Powell, 1983;
Scott, 1989), sociology avoids the requirements for hyper-rationality that has plagued
neoclassical economic thinking. Of course, since institutional economics borrows heavily
from the sociology of institutions and organizations, issues both of bounded rationality and
of power and influence have come to figure prominently in its analyses. They also figure in
Agency theory and in theories of incomplete contracting (Jensen and Meckling, 1976; Hart,
1995; Grossman and Hart, 1988).
The new institutional economics aspires to bridge the gap between neoclassical economics
and organization theory (Williamson, 1985; Furubotn and Richter, 1998). Yet it remains
weighed down by the neoclassical perspective on information. It acknowledges the
existence of friction in the transactional processes that drive the economic system, but
offers little or no theorizing about it. At best, it can differentiate between markets - an
external information environment in which data is well codified and can therefore flow
freely - and hierarchies - an internal information environment in which the flow of data is
viscous on account of the tacit nature of the knowledge involved. The first type of more
analytically tractable environment has typically been the province of economists; the
second, more qualitative environment has been left to organizational theorists.
Perhaps on account of its more qualitative material, organizational sociology has addressed
the problem of knowledge in organizations, but not much that of data or information.
Working in the interpretive tradition initiated by Weber, it has focused on sense-making,
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
the process through which information is interpreted and converted by receivers into
intelligible knowledge (Weick, 1995; Daft and Weick, 1984; Gioia and Chittipeddi, 1991).
But how the codes on which information is borne come into being in the first place is a
question that needs to be addressed before one can progress on to sense-making.
Habermas's theory of communicative action, for example, sees meaning as something to be
freely negotiated between interacting agents (Habermas, 1987). But can the idea of an open
negotiation realistically apply to the codes that agents inherit and draw upon in their
interactions? Such codes do much to shape the possible meanings that are up for
negotiation. Some of the concepts that organizational sociologists apply to knowledge will
also apply to information7, but for this to yield a credible result, they would have to explore
the nature of data as well as that of information.
4. The Contribution of Information Theory
The discipline that comes closest to doing this is information theory. But, originating as it
does in an engineering tradition, information theory concerns itself primarily with the
challenge of information transmission rather than with problems of information content or
meaning (Nyquist, 1924; Hartley, 1928; Shannon, 1948). It is more abstract in its approach
to information than is sociology, being concerned with the technical characteristics of
communication channels independently of the nature of message sources, senders,
receivers, or message destinations. It seeks to establish efficient encoding strategies for
channels subject to noise.
By relating the definition of information to the freedom of choice we enjoy when we choose
a particular message from among all the possible messages that we might transmit, it
becomes possible to calculate the amount of information carried by that message. It turns
out to be the inverse of its probability of appearance. Since within the framework provided
by information theory, any message consists of a sequence of symbols drawn from a given
repertoire of symbols, the theory allows one to assess the effectiveness of different coding
schemes using different symbolic repertoires in a channel. Shannon’s Mathematical Theory
of Communication (Shannon and Weaver, 1949) yields a number of fundamental theorems
which set theoretical limits to the amount of information that a channel of given capacity is
7 Giddens's theory of structuration, for example, and his concepts of domination, signification and
legitimation (Giddens, 1984) can be used to analyze the distribution of both knowledge and information in a
social system, the nature and extent to which these are codified, and their normative status.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
able to transmit, both in the presence and absence of noise. Whether or not the limit is
reached in a particular situation will turn on the choice of symbolic repertoires and syntactic
rules, as well as on the choice of coding scheme.
The amount of information that can be transmitted, then, is a function of the size of the
available repertoire of distinct symbols or states that is available, the relationships between
symbols – i.e., the syntax - as well as the degree of fidelity required given the amount of
noise in the channel. Information theory is primarily concerned with maximizing the
fidelity of transmission at an acceptable cost – Shannon and Weaver (1949) refer to this as a
technical level problem (level 1). As Shannon took pains to point out in his 1948 paper,
information theory is not particularly concerned with what the symbols actually mean - a
semantic level problem (level 2) - or with whether a given message has the desired effect on
a given message destination—an effectiveness level problem (level 3). These he viewed as
problems to be addressed by social scientists rather than engineers. Shannon thus sought to
offer a clear line of demarcation between information and knowledge.
Crucially, information theory takes the repertoire of symbols to be transmitted as a given. It
does not ask how the repertoire came into being, whence the distinctness of the symbol
system came from, or whether the symbolic repertoire was established by prior convention
or through a gradual process of discovery. Yet, before we are in a position to extract
information from a symbol, we first need to extract the information that it is indeed a
symbol and hence an acceptable candidate for further processing. It must, therefore, be
distinguished from other stimuli that might register with an agent as data In short,
information theory ignores the question of data, of how a given repertoire of symbols – a
pre-selected collection of states – gets itself registered with an agent as a data set from
which information can then be extracted8.
If, as we have argued, data is a discernible difference between two states, at a purely
theoretical level, the limiting case of what constitutes a difference is given by the calculus.
It defines, in the limit, what can ever count as data. Perhaps the physically limiting case of
data is given by Planck's constant, which defines the smallest discernable event that can
8 Interestingly, Blackwell applied the precepts of information theory to states rather than symbols. These could
then acquire the status of commodities in an Arrow-Debreu analytical framework. Blackwell’s work was to
influence game-theoretic and other work on the economics of information (Blackwell and Girschik, 1954;
Lipman, 1991; Plott and Sunder, 1978; Geanakoplos, 1992; Rubinstein, 1998).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
pass off as a state. But for us as sentient beings, what counts as data is what we can actually
discern. Our ability to discern differences between states only operates within a certain
physiological range (Hargreaves Heap and Varoufakis, 1995). Outside that range, we
cannot be sure that the different states that constitute our data are orthogonal to each other
and hence capable of yielding a viable repertoire, as required by Shannon.
Data, then, and the regularities that reside within the data, are properties of events and
things “out there” in the world – i.e., physical processes and products - that become
available to us as sentient beings through our physiological apparatus, often amplified by
instruments and other artefacts. Information, by contrast, is relational. As Bateson put it, it
is "the difference that makes a difference"—and that means making a difference to someone
(Bateson, 1971). Thus we might say that regularities within data, an objective property of
events and things, convey more or less information to different individuals, depending on
their respective circumstances, such as their individual histories, their values and emotional
make up (Damasio, 1999), their mental models, and the specific state of their expectations
at any given moment.
The early founders of modern information theory—Nyquist, Hartley, Shannon—imported
from thermodynamics the concept of entropy, which Shannon then associated with the
amount of information H gained in a message. Building on the concept of entropy that
information theory shares with thermodynamics, we would like to suggest that information-
bearing data may be likened to free energy in a physical system. That is to say, data that
carries information retains a capacity to do work – i.e., it can act on an agent's prior state of
expectations and modify it. Data that carries no information may be likened to bound
energy in physical systems: to the extent that it leaves an agent's state of expectations
unmodified, it has performed no work on its expectational structure.
Note that we are dealing here with both an objective term—the quantity of information that
can potentially be carried by a given data set9 –and a subjective term - the amount of
information that can be extracted in practice from the data set by a situated agent. When we
claim that information is relational, it is with respect to the second of these terms. This
"subjectivist" view of information, however, based as it is on an agent's situated
expectations, confronts the "objectivist" view of information developed by Shannon, one
9 This quantity has been calculated for different states of physical matter by Seth Lloyd (Lloyd, 2000).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
that is based on conventionalized expectations. The English language, for example, contains
an objective amount of information based on the relative frequency of appearance of letters
of the alphabet and of certain types of words, such as articles and pronouns. In Shannon's
view, information content is set by the ratio of actual to possible events. In the examples
that he gives, however, both the repertoire of possible events and their frequency are fixed a
priori, so that the computation of information content is straightforward. Yet, to stay with
the example of the English language, as soon as we move up to the level of sentences and
paragraphs, the number of possible sentence constructions moves to infinity. Does this
mean that information content moves to infinity? No, simply that the repertoire of possible
sentences is now largely shaped by the circumstances in which any given sentence is likely
to be uttered – i.e., by its context. But context varies in the extent to which it is shared
across individuals. Some contexts will be unique to individuals, while other contexts will
be widely shared.10
Native English speakers, for example, will share almost identical expectations concerning
the frequency of appearance of letters in English words. They will share somewhat similar
expectations concerning the frequency of appearance of many classes of words in a well-
constructed English sentence. They will share far fewer expectations, however, concerning
the rate at which other words appear in the sentence, for these will depend on particular
circumstances. The discourse that might take place in a biology laboratory, for example,
will be meaningful to a much smaller group of people than the one taking place on a
televized chat show. In sum, it is shared context, the generator of inter-subjective
objectivity (Popper, 1959) that stops information content from ballooning to infinity and
that renders discourse possible.
Shannon takes care of this difficulty largely by avoiding it. Given his focus, this was not
unreasonable. As a communication engineer, he was concerned mainly with the objective
and computable aspects of information and the requirements that these might impose on a
communication channel. Thus Shannon addressed what he called the level 1 or technical
problem (was the message received the same as the message sent?) and confined his
analysis to well defined and delimitable repertoires. What he called the level 2 or semantic
10 Information in the objectivist view can be seen as the higher bound of the ensemble of all possible
“subjectivist” or “inter-subjectivist” interpretations that could be extracted from the data. Yet in any but the
most simple contexts, the objectivist view confronts a Godelian ‘undecidability’ problem .
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
problem (is the received message understood?) was not his concern. This depended on
whether the receiver possessed the relevant code – i.e., some familiarity with the alphabet,
the vocabulary and the syntactic rules of the English language, etc. Note that, even here, the
repertoire was assumed by Shannon to be closed: the alphabet is limited in size as are both
the vocabulary and the syntactic rules that have to be attended to. Finally, what Shannon
called the level 3 or effectiveness problem (does the message lead to the desired
behaviour?), was completely outside his purview. Both levels 2 and 3 we identify with
knowledge.
It is clear that, where symbolic repertoires and syntactic structures are established by
convention rather than by discovery, technical level (level 1) communication issues need
not concern themselves with the idiosyncratic characteristics of communicating agents.
However, the minute we move to the semantic level (level 2) or to the effectiveness level
(level 3), the dispositional states of the agents - i.e., their prior knowledge - become
relevant. Agents are situated processors and transmitters of data. The individual agent's
memories as well as his preference orderings - and hence values and emotional dispositions
(Damasio, 1999) - therefore need to be reckoned with. It is at levels 2 and 3, then, that the
idiosyncrasy and potential subjectivity of context becomes most manifest. Here, selection is
constrained less by rules than by personal style and preference.
We can represent the issue that we are discussing with a diagram In the rectangle of figure
2, we variously mix expectations—and hence probabilities—based on agreed conventions
concerning what constitutes an event, the number of recurrences of that event that constitute
a fair sample, etc., with expectations based on personal experience. The first type of
probability will lend itself to a frequency interpretation whereas the second will lend itself
to a Bayesian or subjectivist interpretation. We subdivide the rectangle into three zones and
associate each zone with one of Shannon's three levels. We see that Shannon's level 1
problem—the technical problem—leaves little or no scope for the subjectivist approach to
probability. It is also the level that is the most computationally tractable and the one to
which Shannon himself decided to confine his analysis. His level 2 problem—the semantic
problem—is one that offers somewhat more scope for subjective probabilities to kick in. In
language, for example, syntactic constraints and word usage will conventionalize
expectations to some extent, but personal idiosyncrasies and style will inject a strong
subjective element into the communication process. Finally, Shannon's level 3 problem—
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
the pragmatic problem—leaves little scope for the frequency perspective, since at this level,
conventions hardly appear as anything other than subjectively experienced and highly
variable constraints.
Fig. 2. Frequency and Bayesian interpretations in Communication
The implication of the above is that, whatever intrinsic regularities it contains, to the extent
that data only carries information when it can modify an expectation, what constitutes data
tout court for me, might turn out to be information-bearing data for you. How far we are
aligned in our information-extraction strategies will depend on how far our respective
expectations are shaped by conventions, that is, socially shared encoding rules and
contextualizing procedures, or by idiosyncratic circumstances - codes and contexts that are
not widely shared. The act of extracting information from data constitutes an interpretation
of the data. It involves an assignment of the data to existing categories according to some
set of pre-established schemas or models that shape expectations. For this to be possible,
such schema or models must already exist in some form or other.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
But how do such schemas and models come into existence in the first place? They do so
primarily through explicit or tacit rules of inference. Explicit rules will for the most part be
applied to codes; implicit rules will be applied primarily to context. Expectations and
categories co-evolve, with expectations shaping the categories that we create, and these,
once created, in turn shape the evolution of subsequent expectations11. Our categories
condition the dispositions that we adopt towards the world – i.e., our knowledge, taken here
in the Popperian sense of a disposition towards action (Popper, 1983)12. Thus, data can only
constitute information for an agent who is already knowledgeable. Data can be viewed as a
low energy system that acts informationally rather than mechanically (Boisot, 1995), that is
to say, it gives rise to intentional action rather than mere mechanical movement13. Guided
by the structure of its expectations, an agent first extracts what constitutes information for
him from the regularities available in a data stream and then acts upon it (see Figure 1).
Given its almost exclusive focus on the technical level of communication, the work of
information theory has largely ignored such issues. These, occurring as they do at Shannon
and Weaver’s level 2 and their level 3 – i.e., at the level that we have identified with
knowledge rather than information - have proved to be of more interest to interpretative
sociology. In organizational sociology, the semantic problem shows up as a concern with
sense-making (Weick, 1995) or bounded rationality (Simon, 1945; Kahneman and Tversky,
82), whereas the pragmatic problem shows up as a concern with power, values, and
influence (Habermas, 1987). Level 2 and level 3 problems are also of relevance to
institutional theory (DiMaggio and Powell, 1983; Scott, 1989). It is clear, however, that at
these levels, we are far removed from an economic world in which agents can be assumed
to have common knowledge of rationality, a consistent alignment of beliefs, and rational
expectations (Aumann, 1976). But if information theory’s concern with bits and bytes led it
to shun the issue of knowledge, it also managed to sidestep the issue of data by assumption:
for all intents and purposes, information and data were the same thing. Although it never
explicitly claimed otherwise, almost unwittingly the new discipline of information physics
has highlighted the issue.
11 Kantians believe that the categories came first, while Lockeans believe that expectations came first and that
these were shaped inductively by the recurrent features of our experiences. The debate continues; mercifully,
we need not get involved.
12 Kenneth Arrow has the same expectational view of information as Popper does. See Arrow (84).
13 This is not to say that both informational and mechanical effects cannot be present at the same time. But
where energy acts informationally, we can affort to ignore its mechanical effects on behaviour. These are
negligible.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
5. The Physics of Information
According to the late Rolf Landauer, "Information is physical" (Landauer, 1999), and the
most fundamental analysis of the nature of information so far carried out originates in
physics. Even within physics itself, since the most fundamental analysis of physical
processes takes place at the quantum level, it is within the new field of quantum information
theory that we confront the deepest level of analysis of information An important
breakthrough for the development of quantum information theory was the discovery that
quantum states could be treated as if they were information (Nielsen and Chuang, 2000).
Thus, if information is physical, what is physical is also information (Lloyd, 2000).
Quantum information theory, being broader in scope than classical information theory,
operates at the most abstract level, quite removed from any social science conception of
information. Can such a view of information have anything to offer the social sciences?
If information is physical, then, like any other physical process, it is subject to the second
law of thermodynamics. The physical entropy involved here, however, must be
distinguished from the Shannon entropy, even though the two are closely related. One
might, in effect, say that Shannon entropy is predicated upon thermodynamic entropy. In a
closed system, both types of entropy-generating processes turn out to be irreversible14.
Although physicists have not much concerned themselves with it, the distinction that we are
drawing between data, information and knowledge is implicit in the work being done in the
Physics of Information (Zurek, 1990; Feynman, 1996; Feynman, 1999; Bennett, 1999;
Landauer, 1999). If the bit is the fundamental unit of analysis in classical information
theory, then the qubit is the fundamental unit of analysis in quantum information theory.
Just as a classical bit is in one of two possible states, 0 or 1, so a qubit has two possible
eigenstates |0 or |1. One difference between a bit and a qubit, however, is that the latter
can also be in any well-defined linear combination of the two eigenstates, |0 or |1. Another
difference is that, whereas we can directly examine a bit to determine what state it is in, we
cannot directly examine a qubit to determine its quantum state without destroying that state.
In short, the eigenstates of the qubit are not available to us as data.
14 It turns out that, at the quantum level, not all computations are irreversible (Bennett, 1999).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
By the postulates of quantum mechanics, any measure that we perform on a qubit reduces it
to one of its eigenstates. This dichotomy between the state of the qubit and what we can
observe lies at the heart of quantum information and quantum computation. At the quantum
scale, we can ask: how much information does a qubit represent in the absence of
measurement? It turns out that nature holds a great deal of "hidden information" in this way,
and it grows exponentially with the number of qubits15.
In the classical world, we assume that we can distinguish, at least in principle, between
different states, since this is what qualifies them as data. Yet in the quantum world, we have
to abandon such an assumption, for, unless the orthogonality between two given states can
be maintained, one can no longer readily distinguish between them and register such a
distinction as data. Without data one cannot extract reliable information from the system
concerning such states. It turns out that, below a certain scale known as the Planck scale,
the orthogonality between two states can no longer be securely established. There are thus
physical limits to our access to data and hence to our ability reliably to extract information
from data.16
These limits first appeared in 1867 in the field of thermodynamics in the shape of
Maxwell's Demon, a microscopic creature that appeared to violate the second law by using
information to distinguish between fast- and slow-moving particles and hence to throw
dissipative processes into reverse (Leff and Rex, 1990). To understand the nature of the
thermodynamic limits on our access to data, we can revert to our earlier and perhaps
somewhat oversimplifying analogy, taking data as corresponding to the general category of
energy, and information-bearing data as corresponding to free energy - i.e., it has a capacity
to do work in the sense that it can modify our expectations, and, hence, the state of our
knowledge. Noise would then correspond to bound energy: it either consists of data that
carries no information for us and can therefore do no work, or it consists of states that
cannot be distinguished from one another and that hence do not even graduate to the status
15 Some, notably Penrose, have argued that quantum effects are also manifest in human cognitive processes
(Penrose, 1994; Green, 2000). We cannot, however, observe each other's mental states directly without
disturbing these states; we can only observe the behavioural outputs of these states.
16 That there are biological limits to our access to data as well has been know since the work of Fechner in the
nineteenth century.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
of data for us. Noise cannot modify our expectations. Like bound energy, it can perform no
work.
Now, although knowledge itself is dispositional, it reveals itself in purposeful agent
behaviors such as data processing, data transmission and actions based on these. We
hypothesize that data storage, on the one hand, and data processing, transmission and
purposeful action, on the other, can be respectively likened to the build up of potential
energy and to its subsequent exploitation as kinetic energy. As a disposition to act, then,
knowledge corresponds to potential energy - a stock; and as purposeful action or behavior,
knowledge corresponds to kinetic energy - a flow17. In open systems, both the
transformation of potential energy into kinetic energy, and the transformation of the latter
into work, are subject to dissipation.
Landauer (1990) demonstrated that energy dissipation occurs both in the information
storage process as well as in the information transmission process as a result of information
erasure. No information is erased in a reversible computation, however, because the input
can always be recovered from the output. When we say, therefore, that a computation is
reversible, we are really saying that no information is erased during the computation.
Landauer's principle provides the link between energy dissipation and irreversibility in
computation, stating that, in order to erase information, it is necessary to dissipate energy.
The principle can be stated thus (Landauer, 1990):
If a computer erases a single bit of information, then the amount of energy dissipated into the
environment will be at least kbT ln 2, where kb is a universal constant known as Boltzmann's constant,
and T is the environmental temperature of the computer.
The laws of thermodynamics also allow us to express Landauer's principle in terms of
entropy rather than in terms of energy and dissipation (Landauer, 1990):
If a computer erases a single bit of information, then the environmental entropy will increase by at
least at least kb ln 2, where kb is Boltzmann's constant .
17 There are, of course, important differences between knowledge and potential energy, on the one hand, and
behaviour and kinetic energy, on the other. A stock of knowledge is not depleted in the way that a stock of
potential energy might be. It can only be dissipated, in the sense that the information structures that constitute
the stock are gradually eroded. Likewise, behaviors, through the mechanism of learning can actually build up
a knowledge stock rather than depleting it - which is what kinetic energy does to potential energy. Clearly,
reasoning by analogy must know its limits.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
It should be noted that Landauer's principle effectively provides us only with a lower bound
on the amount of energy that must be dissipated to erase information. Clearly, if all
computational processes were reversible, then the principle would imply no lower bound on
the amount of energy dissipated, since no bits would in fact be dissipated during
computation (Nielsen and Chuang, 2000).
Maxwell's Demon is located at the meeting point of a physics of energy and a physics of
information. Bolzmann's definition of entropy links the two types of physics. An
informational limit is reached under two quite different conditions. The first occurs when
the energy expenditures incurred by data capture and transmission activities required to
distinguish between two states, and hence to create discernable data - often performed by
specialized equipment - itself exerts a mechanical effect on those states, thus preventing
them from stabilizing enough to get themselves detected - Heisenberg's uncertainly
principle describes this condition. The mechanical effects of such energy expenditures then
swamp and overwhelm their informational effects. The second occurs when the Demon
needs to store transmitted data in memory for subsequent processing. Assuming that the
Demon's memory is finite - i.e., it is subject to bounded rationality (Simon, 1945) - it will
sooner or later confront the need to erase stored data in order to make way for new data.
Landauer's principle tells us that at that moment data will be lost and entropy levels - both
thermodynamic and informational - will increase. However, thermodynamic entropy and
information entropy are quite distinct from one another. Although both draw on
Boltzmann’s formula, the first refers to the regularities or lack of them in discernible states-
of-the-world – that is, in data - whereas the second refers to the information that can be
extracted from such states by a knowledgeable observer. In sum, if social scientists conflate
information and knowledge, physicists conflate data and information. In the next section, by
means of a simple diagram, we will indicate why both types of conflation matter.
6. An economic interpretation of the principle of least
action
Any physical system is subject to the principle of least action, an integral variational
principle initially put forward by Maupertuis in 1744 that establishes the difference between
the actual motion of the system and all of its kinematically possible motions during a finite
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
time interval (Barrow and Tipler, 1986). According to Green, when the observables of the
system, such as its energy, its momentum, its angular momentum, its central vector and
certain other charges, have prescribed values on the boundary of any region of space and
time, they will vary in such a way that the total action within the region has its minimum
value (Green, 2000). This will be as true of dissipative systems as it will be of Hamiltonian
systems. To the extent that the system has a capacity for storing memories of earlier states -
and this does not require that the system be intelligent or even alive - then it will be able to
use data and information in such a way as to minimize the action.
Being universal in scope (Omnes, 1999), the principle of least action implies that nature as
a whole makes choices that are economic in their outcomes18. How might Maxwell's
Demon apply the principle? In effect, it allows us to posit the existence of a trade-off
between the Demon's consumption of energy and his consumption of data resources as it
attempts to sort out fast-moving from slow-moving particles. Such a trade-off can usefully
be represented by means of a scheme that is somewhat reminiscent of a production
function, but the purpose of which is limited to illustrating the economic nature of the
principle of least action.
A production function is a schedule showing the maximum amount of output that can be
produced from any specified set of inputs (Fergusson, 1969). In neoclassical production
functions that take capital and labor as inputs, information and knowledge are not explicitly
represented as factors of production in their own right, although, in talking of capital and
labor, we may take them to be implicitly present. The knowledge embedded in machinery
and equipment, for instance, clearly forms part of the capital factor, and the labor factor
clearly embodies the know-how and experience of employees. Given that in the so-called
“new economy” information and knowledge have clearly moved center-stage, some have
claimed that they should therefore become critical productive factors in their own right
alongside capital and labor (Bell, 1973; Romer, 1986; Romer, 1990)19.
Yet, given that information and knowledge are already implicitly embedded in traditional
productive factors, this would result in double counting. As an alternative, therefore, one
18 In the nineteenth century, this was referred to as the economy of nature.
19 By the term new economy, we mean more than an economy driven by the Internet phenomenon. We
therefore avoid having to take sides in the current debate as to whether there is in fact a new economy.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
could move up to a more abstract and general level and bring together two different classes
of productive factors: 1) purely physical factors, such as space, time, and energy – these
would be measured in physical units such as meters, seconds, and joules, and 2) data
factors, being discernible differences in the states of the physical factors – these would be
measured in bits (Boisot, 1995; Boisot, 1998). Note that, in this new scheme, information
and knowledge are not taken as factors of production at all. According to our earlier
arguments, information constitutes an extraction from the data factor that results in
economizing on that factor and hence in a move towards the origin. Knowledge, likewise,
economizes on data-processing - and hence on the consumption of data inputs - more so in
the case of abstract knowledge than of concrete knowledge20.
An example of the new scheme is shown in figure 3. As will shortly be apparent, much as it
may look like one, it is not actually a production function. In the diagram, we can
distinguish two types of movement, one along isoquants and another across them. A move
to the left along an isoquant represents a progressive substitution of data for physical
factors, something that happens when, by gradually accumulating the data of experience,
systems "learn-by-doing", with less expenditure of time, space, and energy, in whatever
task they are performing – manufacturing aircraft wings, miniaturizing electronic
components, etc. Learning-by-doing can only work for systems that can store past states –
i.e., for systems that have memory. Some purely physical systems have memory and all
living systems do. By implication, a move to the right along an isoquant can be interpreted
either as forgetting, an erosion of memory, or as the workings of bounded rationality. Both
rightward and leftward movements are possible. A downward vertical movement across
isoquants and towards the origin in the diagram, by contrast, represents the generation of
insight, the extraction of information from data to create new, more abstract knowledge
concerning the structure underlying phenomena. This second movement – the joint effects
of pattern recognition and computational activities – is discontinuous, reflecting the
unpredictable nature of creative insights (Miller, 1996). It makes it possible to reach the
same output levels as before with less data processing and hence a lower consumption of
data inputs. In addition to having memory, a system that has a capacity for insight must also
be intelligent, that is, it must be capable of processing data in order to extract information
from it in the form of patterns or structures.
20 Ernst Mach’s “Principles of the economy of thought” were an important source of inspiration for Hayek
(Mirowski, 2002).
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
Fig. 3. Data vs. physical factors scheme
Our scheme and the neoclassical production function have some similarities. For example,
they both take movement along an isoquant as representing technical change – i.e., a
change in the mix of data and physical resources that generate a given output - and
movement across isoquants towards the origin as representing technical progress (Boisot,
1998) – i.e., a reduction in the quantity of data and/or physical resources required to
generate that output. Yet the two schemes differ in three important ways.
First, while the neoclassical production function offers no preferred direction for
movements along an isoquant, the broad tendency to substitute data factors for physical
factors in our scheme – a process of variation, selection and retention that results in data
accumulating in the form of memory — imparts a direction to technical change and to
technical progress. Why? The clue resides not in the evolutionary nature of knowledge –
although this is certainly a factor – but in the evolutionary nature of agents, individual or
corporate (Metcalfe, 1998). To the extent that evolution enhances both the memory and the
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
data processing capacity of agents – this can be achieved either via biological evolution or
via the artefacts of cultural evolution (Boyd and Richerson, 1985; Clark, 1997) - they are
able to make better use of whatever data accumulates over time, and this at a lower cost
than that of using the physical resources available to them. Thus, if intelligence is selected
for by evolution, intelligence, in turn, will demonstrate a selection bias in favor of data over
physical resources
In effect, in contrast to the neoclassical production function in which movement along an
isoquant is reversible, the arrow of time is at work in our scheme, allowing it to describe
irreversible and hence path-dependent processes that, according to circumstances, might be
characterized as being either evolutionary or as developmental. We must emphasize that the
arrow of time manifests itself in global rather than local behaviors. The general tendency
for a leftward movement up an isoquant is likely to have many local exceptions - brought
about either by forgetting or by bounded rationality – that move it in the opposite direction.
Second, our scheme is able to account for technical progress. Although in both the
neoclassical and in our scheme, technical progress in described by a jump across isoquants
towards the origin, in the neoclassical case, such a discontinuity cannot be explained; it had
to be exogenously given. In our scheme, by contrast, a discontinuous jump from one
isoquant to another is accounted for by a discontinuous jump in a living system’s own
learning processes – i.e., it is accounted for by the discontinuous phenomenon of insight,
the extraction of informative patterns or gestalts from data, and their subsequent conversion
into knowledge.
Third, the data and physical factors that make up our scheme present quite distinct
economic properties. While, in the neoclassical production function, purely physical factors
are naturally subject to scarcity and hence appropriable, the data factors of our own scheme
are not. While they may not always be immediately accessible—and in that sense they may
be considered scarce - once one has secured them, they can often be replicated and
distributed at almost zero marginal cost. Providing it has found the right physical substrate,
therefore, data will propagate rapidly and extensively. Scarcities will then only appear in
the form of a living system's limited capacity to receive, store, process, and transmit data,
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
not in the data factors themselves21. For this reason, data factors are much more difficult to
appropriate and to subject to traditional forms of economic exchange than purely physical
factors. They are hard to price and this makes it hard to use price signals to guide a
substitution of data factors for physical ones. Our scheme can illustrate such a substitution
process; it cannot analyze it.
To summarize: except, perhaps, for the universe as a whole, there are no perfectly closed
systems in the real world. Open systems are prey to unwanted interactions with their
environment that get registered as noise when viewed informationally. Economics has
tended to ignore the implications of the fundamental openness of the systems they study.
Our scheme, by allowing the representation of the effects of time and entropy in the
economic process, rectifies the situation. Once you admit learning, development and
evolution into the picture, you admit irreversible processes. But our scheme also suggests
that the entropy concept is but one side of the coin when dealing with the second law of
thermodynamics. Irreversible processes can lead to emergent, order-creating outcomes as
well as to entropic ones, those that allow living things to jump across isoquants and move
towards the origin in pursuit of factor savings (Brooks and Wiley, 1988). Although we may
agree with Shapiro and Varian when they observe in Information Rules (1999) that the
information economy has not yet repealed the laws of economics, we feel that it poses
explanatory challenges to economics—well captured by the way that novelty and new
knowledge emerges in living systems and organizations—that the discipline has yet to take
on board.
7. Implications
We can briefly summarize our discussion in the following three propositions:
21 We can see the logic of our new scheme at work in the way that organizations are today attempting to
handle large amounts of transactional data. Data mining, for example, is the process of extracting information
from data. People will not pay for data, but as information extraction becomes ever more difficult and user-
specific - i.e., customized - people will pay for information. To the extent that data can be turned into
information that has relevance for someone, that someone will in principle be willing to pay for the data
processing and transmission economies on offer. What such economies offer is the possibility of reallocating a
key data processing resource possessed by all intelligent agents in finite quantities: attention. An information
economy, by implication, has to be an attention economy as well.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
1.Information is physical (Landauer, 1999). It is a constituent element of all physical
processes and hence cannot treated as something epiphenomenal to the economic process. It
must be engaged in on its own terms.
2.Economic agents subject to the principles of least action and to the effects of the second
law of thermodynamics aim to economize on their consumption of both physical and data
resources by deploying effective cognitive and behavioral strategies.
3.Effective cognitive strategies extract information from data and then convert it into
knowledge. Effective cognitive and behavioral strategies vary from agent to agent as a
function of their situation, of their prior individual knowledge, of their values, and of their
emotional dispositions.
What follows from our three propositions?
Developing further the difference between data, information and knowledge, data generates
thermodynamic entropy, which we shall label entropy 1. It involves the erasure of
differences between physical states. Information, by contrast, generates Shannon entropy,
which we shall label entropy 2. It involves the erasure of differences between symbols. The
difference between physical states might well be maintained, but the form given to such
states no longer yield unambiguous symbols. Finally, knowledge generates cognitive
entropy, which we shall label entropy 3. It involves the erasure of differences between the
possible contexts required for the interpretation of either states or symbols.
All these different types of entropy constitute variations on Bolzmann's formula
N
pi log pi. for i = 1…n,
where n describes either the number of possible data states, the number of symbols in a
repertoire, or the number of interpretative contexts that are compatible with a given set of
states or symbols. N gives the message length, and we hypothesize that an inverse
relationship exists between N and n. Efficient coding, however, should reduce both N and n
to the extent that is builds on correlations between states, symbols, or interpretative
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
contexts. Where such correlations are not given a priori, they must be discovered. In the
absence of memory, however, an agent has no way of discovering such correlations so that,
in effect, n can now potentially increase without limit. This makes Boltzmann’s formula
meaningless since it cannot be used as a basis for stable expectations.
Entropies 1 and 2 are to be found at Shannon's technical level. Entropy 3 is to be found at
Shannon's semantic and effectiveness levels. At the semantic level, it can occur because the
receiver does not know the codes or what, specifically, they refer to—this, in effect, is
context narrowly defined—and at the pragmatic level it can occur because the receiver does
not know to embed the message as a whole into an appropriate context. Entropy 1 has the
effect of increasing Entropy 2 and Entropy 3. However, redundancy at the semantic and
effectiveness levels can mitigate the effects of entropy 1.
Economics at best has only ever operated at Shannon's technical level. By largely ignoring
problems of meaning and values, it has only scratched the surface of Shannon's semantic
and effectiveness levels. Yet the implication of our analysis is that, strictly speaking, there
is no such thing as common knowledge and there is common information only to a limited
extent. Only data can ever be completely common between agents. As Metcalfe puts it,
agents may live in the same world, but they see different worlds (Metcalfe, 1998). In its
treatment of information, economics thus fell between two stools. On the one hand, it
eschewed the complexities of the "soft" approach to knowledge and information associated
with the semantic and effectiveness levels - and with the social and cognitive sciences as a
whole. On the other hand, it never really dug into the foundations the way that physics did,
in order to distinguish entropies 1 and 2 from one another22. It therefore allowed the concept
of information to take whatever form was needed to maintain analytical and
computationally convenience.
8. Conclusion
From "soft" sciences such as sociology right across to "hard" sciences such as physics,
information has become a central concern. Economics, however, has tended to treat
information as something unproblematic, an auxiliary concept that can be left largely
22 Yet if the physics of information helped effectively, to distinguish entropy 1 from entropy 2, this did not
result in a distinction within the discipline between data and information.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
unanalyzed. Yet, in post-industrial economies, information has now become the main focus
of economic transactions, and not merely a support for them. Economists, therefore, cannot
afford the luxury of neglecting the conceptual foundations of an economics of information
in this way.
Shapiro and Varian have argued that the laws of economics apply to the information
economy no less than to the energy economy that preceded it. This is undoubtedly true and
certainly needed to be said. The issue, however, is not about the applicability of economic
laws, but about their scope. The physics of Newton was not displaced by that of Planck and
Einstein, rather it ended up having to share the stage with them. Likewise, the physics of
information neither falsifies the economic laws of the energy economy, nor does it render
them irrelevant. What it brings out, however, is that, given their limited engagement with
the concept of information, such laws will have trouble dealing with many of the
phenomena associated with the evolution and growth of knowledge in general and with the
emergence of the new economy in particular. They therefore need to be complemented with
more encompassing and general laws that take into account the pervasive roles played,
respectively, by data and information in all physical processes, as well as that played by
knowledge in biological ones. Such roles are distinct and complementary and in need of
clear articulation. This paper has attempted to provide some initial theoretical reflections on
a task that still lies ahead.
References
Arrow K J (1984) The Economics of Information. Belknap Press of Harvard University
Press, Cambridge, MA.
Arrow K J (1984) On the agenda of organizations. In: Arrow K J (ed) The economics of
information 1st edn. pp 167-184. Belknap Press of Harvard University Press,
Cambridge, MA.
Arrow K J (1996) The Economics of Information: An Exposition. Empirica 23 : 119-28
Aumann R J (1976) Agreeing to disagree. Annals of Statistics 4 : 1236-1239
Barrow J D, Tipler F J (1986) The Anthropic Cosmological Principle. Oxford University
Press, Oxford
Bateson G (1971) The Cybernetics of 'Self': A Theory of Alcoholism. Psychiatry 34 (1):
1-18
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
Bell D (1999) The Coming of Post-Industrial Society. Basic Books, New York
Bennet C H (1999) Quantum Information Theory . In: Hey A J G (ed) Feynman and
Computation: Exploring the Limits of Computers pp 177-190. Perseus, Cambridge,
MA.
Binmore K (1990) Essays on the Foundations of Game Theory. Basil Blackwell, Oxford
Blackwell D, Girschik M A (1954) Theory of Games and Statistical Decisions. Wiley, New
York
Boisot M H (1995) Information Space: A Framework for Learning in Organizations,
Institutions and Culture. Routledge, London
Boisot M H (1998) Knowledge Assets: Securing Competitive Advantage in the
Information Economy. Oxford University Press, New York
Boyd R, Richerson P J (1985 ) Culture and the Evolutionary Process. The University of
Chicago Press, Chicago
Boyle J (1996) Shamans, Software and Spleens: Law and the Construction of the
Information Society. Harvard University Press, Cambridge, MA.
Brooks, D., and Wiley, E., (1986) Evolution as Entropy: Toward a Unified Theory of
Biology, Chicago: The University of Chicago Press
Bruner J S (1974) Beyond the Information Given: Studies in the Psychology of Knowing.
George Allen and Unwin, London, UK.
Clark A (1997) Being There: Putting Brain, Body, and World Together Again. The MIT
Press, Cambridge, MA.
Crary J (1999) Suspensions of Perception: Attention, Spectacle, and Modern Culture. The
MIT Press, Cambridge, MA.
Daft R, Weick K E (1984) Towards a Model of Organizations as Interpretation Systems.
Academy of Management Review 9 (2): 284
Damasio A R (1999) The Feeling of What Happens: Body and Emotion in the Making of
Consciousness. Hartcourt, San Diego
Deleuze G (1969) Logique Du Sens. Les Éditions de Minuit, Paris
Derrida J (1967) L'Écriture Et La Différence. Les Éditions du Seuil, Paris
DiMaggio B, Powell W W (1983) The Iron Cage Revisited: Institutional Isomorphism and
Collective Rationality in Organizational Fields. American Sociological Review 82
: 147-160
Durkheim E, Mauss M (1963) Primitive Classification. Cohen and West, London, UK.
Fergusson, C.E., (1969) Microeconomic Theory, Homewood, Illinois: Richard D. Irwin
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
Feynman R P (1996) Feynman Lectures on Computation. Addison-Wesley, Reading, MA.
Feynman R P (1999) There's Plenty of Room at the Bottom. In: Hey A J G (ed) Feynman
and Computation: Exploring the Limits of Computers pp 63-76. Perseus,
Cambridge, MA.
Fransman M (1998) Information, Knowledge, Vision and Theories of the Firm. In: Dosi G,
Teece D J,Chytry J (eds) Technology, Organization, and Competitiveness:
Perspectives on Industrial and Corporate Change pp 147-191. Oxford University
Press, New York
Furubotn E, Richter R (1998) Institutions and Economic Theory: The Contribution of the
New Institutional Economics. The University of Michigan Press, Ann Arbor, MI.
Geanakoplos J (1992) Common Knowledge. Journal of Economic Perspectives (6): 53-82
Giddens A (1984) The Constitution of Society: Outline of the Theory of Structuration.
Polity Press, Cambridge
Gioia D, Chittipeddi K (1991 ) Sensemaking and Sensegiving in Strategic Change
Initiatives. Strategic Management Journal 12 : 433-448
Gottinger H-W (1983) Coping With Complexity. Reidel, Boston, MA.
Granovetter M (1985) Economic Action and Social Structure: the Problem of
Embeddedness. American Journal of Sociology 91 : 481-501
Green H S (2000) Information Theory and Quantum Physics: Physical Foundations for
Understanding the Conscious Process. Springer, Berlin
Grossman S, Hart O (1988) One Share-One Vote and the Market for Corporate Control.
Journal of Financial Economics 20 : 175-202
Habermas J (1987) The Theory of Communicative Action I: The Critique of Functionalist
Reason. Polity Press, Cambridge
Hamilton D B (1991) Evolutionary Economics: A Study of Change in Economic Thought.
Transaction Publishers, New Brunswick
Hargreaves Heap S, Varoufakis Y (1995) Game Theory: A Critical Introduction.
Routledge, London, UK.
Hart O (1995) Firms, Contracts, and Financial Structure. Oxford University Press, New
York, NY.
Hartley R V L (1928) Transmission of information. Bell System Technical Journal : 535-
563
Hayek F A (1999) The use of knowledge in society. In: Cortada J W, Woods J A (eds) The
knowledge management yearbook 1999-2000 1st edn. Butterworth-Heinemann,
Woburn, MA
Hirshleifer J, Riley J J ( 1992) The Analytics of Uncertainty and Information. Cambridge
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
University Press, Cambridge, UK
Hodgson G (1988) Economics and Institutions: A Manifesto for a Modern Institutional
Economics. Polity Press, Cambridge
Hodgson G (1993) Economics and Evolution: Bringing Life Back into Economics. Polity
Press, Cambridge
Hodgson G (1999) Economics and Utopia: Why the Learning Economy Is Not the End of
the History. Routledge, London
Hurwicz L (1969) On the Concept and Possibility of Informational Decentralization.
American Economic Review 59 : 513-34
Jensen M, Meckling W (1976) Journal of Financial Economics 3 : 305-60
Jung C G (1971) Collected Works: Vol. 6, Psychological Types. Princeton University
Press, Princeton, N.J.
Kahneman D (1994) New Challenges to the Rational Assumption. Journal of Institutional
and Theoretical Economics 150 : 18-36
Kahneman D,Tversky A (1982) Judgment under uncertainty: Heuristic and biases. In:
Kahneman D, Slovic P,Tversky A (eds) Judgment under uncertainty: Heuristic and
biases pp 3-20. Cambridge University Press, Cambridge, UK.
Koopmans T (1957) Three Essays on the State of Economic Science. Kelley, New York
Kuhn, T.S, (1974) ‘Second Thoughts on Paradigms’, in Suppe, F., (ed), The Structure of
Scientific Theories, Urbana: The University of Illinois Press
Kuhn T S (1996) The Structure of Scientific Revolutions. 3a edn.: The University of
Chicago Press, Chicago
Lamberton D M (1996) The Economics of Communication and Information. Elgar,
Cheltenham
Lamberton D M (1998) Information Economics Research: Points of Departure.
Information Economics and Policy 10 : 325-330
Landauer R (1990) Irreversibility and heat generation in the computing process. In: Leff H
S, Rex A F (eds) Maxwell's Demon: Entropy, Information, Computing pp 188-196.
Adam Hilger, Bristol, UK.
Notes: Originally in IBM J. Res. Dev., 5, 183-91 (1961)
Landauer R (1999) Information is Inevitably Physical. In: Hey A J G (ed) Feynman and
Computation: Exploring the Limits of Computers pp 77-92. Perseus, Cambridge,
MA.
Lange O, Taylor F M (1938) On the Economic Theory of Socialism. University of
Minnesota Press, Minnesota
Lavoie D (1985) Rivalry and Central Planning. Cambridge University Press, Cambridge,
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
UK.
Leff H S, Rex A F (1990) Maxwell's Demon: Entropy, Information, Computing. Adam
Hilger, Bristol, UK.
Lipman B L (1991) How to Decide How to Decide How to... Econometrica 59 : 1105-25
Lloyd S (2000) Ultimate physical limits to computation. Nature 406 (31): 1047-54
MacPherson C (1962) The Political Theory of Possessive Individualism: Hobbes to Locke.
Oxford University Press, Oxford, UK.
Makowski L,Ostroy J (1993) General Equilibrium and Market Socialism. In: Bardhan P,
Roemer J (eds) Market Socialism: The Current Debate Oxford University Press,
New York
Mannheim K (1960) Ideology and Utopia. Routledge and Kegan Paul, London, UK.
Marschak J (1974) Economic Information, Decision and Prediction: Selected Essays, 3
Vols. Reidel, Dortrecht
Marx K (1970) Capital: A Critique of Political Economy. Lawrence and Wishart, London
Mead G H (1962) Mind, Self and Society: From the Standpoint of a Social Behaviourist.
The University of Chicago Press, Chicago, IL.
Metcalfe J S (1998) Evolutionary Economics and Creative Destruction. Routledge,
London, U.K.
Miller, A,I, (1996) Insights of Genius: Imagery and Creativity in Science and Art,
Cambridge, Mass: The MIT Press
Miller G A (1956) The Magical Number Seven, Plus of Minus Two: Some Limits on our
Capacity for Processing Information. Psychological Review 63 (2): 81-96
Mirowski P (2002) Machine Dreams: Economics Becomes a Cyborg Science. Cambridge
University Press, Cambridge, UK.
Muth J F (1961) Rational Expectations and the Theory of Price Movements. Econometrica
29 (3): 315
Myerson R (1999) Nash Equilibrium and the History of Economic Theory. Journal of
Economic Literature 107 : 1067-82
Nelson R R (1994) Economic Growth via the Co-Evolution of Technology and Institutions.
In: Leydesdorff L, Van den Besselaar P (eds) Evolutionary Economics and Chaos
Theory: New Developments in Technology Studies Pinter, London, U.K.
Nelson R R, Winter S G (1982) An Evolutionary Theory of Economic Change. Belknap
Press of Harvard University Press, Cambridge, MA.
Nielsen M, Chuang I (2000) Quantum Computation and Quantum Information. Cambridge
University Press, Cambridge, UK.
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
Nyquist H (1924) Certain factors affecting telegraph speed. Bell System Technical Journal
April 1924 : 324-346
Omnes R (1999) Quantum Philosophy: Understanding and Interpreting Contemporary
Science. Princeton University Press, Princeton
Penrose R (1994) Shadows of the Mind. Oxford University Press, Oxford, UK.
Plott C, Sunder S (1978) Rational Expectations and the Aggregation of Diverse
Information in the Laboratory Security Markets. Review of Economic Studies 45 :
133-53
Popper K R (1980) The Logic of Scientific Discovery. Routledge, London
Popper K R (1992) Realism and the Aim of Science. Routledge, New York, NY.
Prigogine I (1980) From Being to Becoming: Time and Complexity in the Physical
Sciences. W.H. Freeman and Company, New York
Quine W V O (1969) Naturalized Epistemology. In: Quine W V O (ed) Ontological
Relativity and Other Essays Columbia University Press, New York
Romer P M (1986) Increasing Returns and Long-Run Growth. The Journal of Political
Economy 94 (5): 1002-1037
Romer P M (1990) Are Nonconvexities Important for Understanding Growth? American
Economic Review 80 (2): 97
Rosen R (1991) Life Itself: A Comprehensive Inquiry Into the Nature, Origin and
Fabrication of Life. Columbia University Press, New York, NY.
Rubinstein A (1998) Models of Bounded Rationality. The MIT Press, Cambridge, MA.
Scott M F (1989) A New View of Economic Growth. Oxford University Press, Oxford,
UK.
Shannon C E (1948) The Mathematical Theory of Communication. Bell System Technical
Journal 27 : 379-423, 623-656
Shannon C E, Weaver W (1963) The Mathematical Theory of Communication. University
of Illinois Press, Urbana
Shapiro C, Varian H R (1999) Information Rules: a Strategic Guide to the Network
Economy. Harvard Business School Press, Boston, MA.
Simon H A (1997) Administrative Behavior: A Study of Decision-Making Processes in
Administrative Organizations. The Free Press, New York, NY.
Singh S (1999) The Code Book: The Science of Secrecy From Ancient Egypt to Quantum
Cryptography. Anchor Books, New York, NY.
Smith J A (1991) The Idea Brokers. Free Press, New York
Data, information and knowledge: have we got it right?
http://www.uoc.edu/in3/dt/20388/index.html
© 2004 by Max Boisot and Agustí Canals
© 2004 by FUOC
Stiglitz J E (1983) Information and Economic Analysis: a Perspective. The Economic
Journal 95 : 21-41
Stiglitz J E (2000) The Contributions of the Economics of Information to Twentieth
Century Economics. Quarterly Journal of Economics (140): 1441-78
Von Mises L (1969) Socialism: an Economic and Sociological Analysis. Jonathan Cape,
London
Vromen J J (1995) Economic Evolution: An Enquiry into the Foundations of New
Institutional Economics. Routledge, London
Vygotsky L S (1986) Thought and Language. The MIT Press, Cambridge, MA.
Weick K E (1995) Sensemaking in Organizations. Sage, Thousand Oaks
Weimer W, Palermo D (1982) Cognition and Symbolic Processes. Erlbaum, Hillsdale, NJ.
Williamson O E (1985) The Economic Institutions of Capitalism. The Free Press, New
York, NY.
Zurek W H (1990) Complexity, Entropy and the Physics of Information. Perseus, Reading,
MA.
... As the study relies on the content of the published VfM reports, which is in the form of external information, it is paramount to conduct the review and coding of the information with great care. Boisot and Canals (2004) assert that in many cases, individuals tend to lack understanding of the information they have at hand, which contributes to the information not utilized objectively and ultimately has an implication on the extent to which the problems explored are actually dealt with. To address this critical issue and allow for better use of the information, the process of handling the information and capturing its internal features should be taken distinctively from the information itself. ...
... Correspondingly, in this study, we obtain meaning from the information after we formulate the intended categories or data codes. It is worth noting that while information theory places emphasis on the transmission mechanism of the information used (Shannon, 1948), the theory is criticized as not being responsive toward the content and meaning attached to the information (Boisot & Canals, 2004). ...
... To enable proper coding, unequivocal rules should be applied to the codes, while implicit rulings are to be used for the interpretation of the context (Boisot and Canals, 2004). The data utilized in the coding are considered to encompass information only if the agent(s) involved in the creation of the information are recognized as knowledgeable. ...
Full-text available
Article
The study investigates why the value for money (VfM) audit, in its current form, fails to capture the actual state of affairs in the UK public organizations. To address this, we utilize a VfM assessment matrix and key public sector performance indicators to critically evaluate the VfM reports published by two main public bodies in the United Kingdom, that is, the National Health Services and the police authorities, alongside the reports published by the National Audit Office and Her Majesty's Inspectorate of Constabulary and Fire & Rescue Services. Our results reveal that the VfM reports do not clearly show how the 3Es (i.e., economy, efficiency, and effectiveness) associated with the VfM assessment are attained. There are also limited suggestions on the public bodies’ service output or social outcomes and how performance targets are fulfilled. We deduce that the VfM audit's failure to capture these elements significantly curtails the benefits of the VfM exercise to public bodies. We argue for complementing the current VfM assessment with a review of the performance of these bodies based on the services they offer as well as their strategic objectives.
... From an economic point of view, Boisot and Canals (2004) state that too many studies associate data with information, and that "information" as a term is used for both phenomenawhich is why the conceptualization of information in some cases is too broad. For data to become informationthe authors claimeach interpreting unit/actor needs to invest energy into converting it into information that can be interpreted or learned by one or more units/actors, making it valuable (Boisot and Canals, 2004). ...
... From an economic point of view, Boisot and Canals (2004) state that too many studies associate data with information, and that "information" as a term is used for both phenomenawhich is why the conceptualization of information in some cases is too broad. For data to become informationthe authors claimeach interpreting unit/actor needs to invest energy into converting it into information that can be interpreted or learned by one or more units/actors, making it valuable (Boisot and Canals, 2004). In telemedicine networks, the importance of IT and human resources (professionals) can be highlighted in parallel as indispensable input factors in the interpretation of data available in the network and for the continuity of interactions and the exchange and generation of resources. ...
Article
Purpose: Telemedicine, similarly to social media, accelerates information exchange, enriches information, provides better access to information and, furthermore, has an impact on mobilizing resources in business-to-business relationships. This paper aims to contribute to the understanding of the changes brought about by telemedicine, as a new technology, in patient routes. Design/methodology/approach: This case study method was applied to examine five health-care protocols through their patient routes (series of activities) with and without telemedicine technology. The ARA model was applied to examine the changes telemedicine engendered in relation to activities, resources and actors. The strategy of visual mapping was applied for the comparative analysis. Findings: The analyzed cases show that the new resources applied through telemedicine technology modified the number and substance of relevant activities and the set and role of actors who were involved. The quantity or the availability of output information increased in patient routes when new resources were added by telemedicine technology. When technology change occurred, any change in data or information systems – the two building blocks of information – could result in new or modified activities. If data that is used or produced while undertaking an activity change simultaneously along with the information system used for encrypting this data, then this “joint change” will certainly entail some kind of change in the set of activities, resources or actors that are involved. If not, then the activities continued the same as with the face-to-face protocol (without the new technology). Originality/value: The novelty of the paper is that the results highlight the role of information in the extent of change in interactions induced by new technology. Findings about such changes show how information influenced by activities, resources and actors can help decision-makers in relation to the use of telemedicine.
... Um dos pontos fundamentais para se estudar a gestão do conhecimento é compreender o que é o conhecimento em si, e o que o diferencia de dados e de informação. De forma ampla, dados podem ser definidos como padrões de estímulos sem significado, provocados por mudanças de estado no mundo físico e detectados pela capacidade de percepção de um agente (AAMODT; NYGÅRD, 1995;BOISOT;CANALS, 2004). Dados são fatos brutos, medidas e estatísticas, que nada dizem sobre contextos, motivos ou relacionamentos dos acontecimentos que eles reportam. ...
... Um dos pontos fundamentais para se estudar a gestão do conhecimento é compreender o que é o conhecimento em si, e o que o diferencia de dados e de informação. De forma ampla, dados podem ser definidos como padrões de estímulos sem significado, provocados por mudanças de estado no mundo físico e detectados pela capacidade de percepção de um agente (AAMODT; NYGÅRD, 1995;BOISOT;CANALS, 2004). Dados são fatos brutos, medidas e estatísticas, que nada dizem sobre contextos, motivos ou relacionamentos dos acontecimentos que eles reportam. ...
... This research strives to understand the practices of AM designers, through the analysis of their actions and of their underlying selection criteria. To do so, the knowledge they produce and mobilize in a work situation is the start of this work for the following reasons: -Knowledge is personal and resides in individual's head [1]; it is distinguished from information and data [2]. -Knowledge is hard to locate as it involves many dimensions (Fig. 1). ...
Chapter
Since decades, additive manufacturing (AM) is sparking interest in industry and research laboratories. The assets of this process and its associated technologies no longer needs to be proven: part shape freedom, Buy-to-Fly ratio reduction, integration of lattice structures, etc. As these new technologies implies new expertise, new profiles have then emerged in the value chain. Among them, the CAD/CAM (Computer Aided Design & Manufacturing) engineers become a key player. However, these experts still need training as, for some technologies, they still proceed by trial and error, therefore processing rules would merit harmonization. This research strives to understand the practices of AM designers, through the analysis of their actions and their underlying selection criteria. To do so, the knowledge they produce and mobilize in a work situation is the start of this work.This article proposes then a knowledge-based AM approach to represent key concepts and to model action-oriented knowledge. After a state of the art related to knowledge elicitation, elicitation techniques applied to the AM context are explained. A knowledge model applied to Electron Beam Melting process is presented subsequently as a solution.The methodology combines procedural and conceptual knowledge, highlights action rules. It has the benefit of being a dynamic decision-making support for CAD/CAM engineering, as well as modular and easy to update. It could likewise be applicable to manufacturing activities but also to many processes.KeywordsAdditive manufacturing knowledgeRelational knowledge modelAction rules
Chapter
Unlike other well-known business ecosystems, such as the financial markets or service ecosystems, the data marketplace and its ecosystem are immature and the observable interactions are quite limited. To understand the aspects and interactions of data, stakeholders, and other entities in the marketplace, not only is the analysis of the individual pieces of data important but the investigation of the structural characteristics of the marketplace is also important. In this chapter, we attempt to elucidate the value chain of data businesses and stakeholder interactions in the ecosystem, using the structured knowledge for data utilization generated in the IMDJ workshops, the variable-based data networks, and stakeholder-centric data business networks.
Chapter
Human history has been a series of relentless efforts and actions of accumulating knowledge by finding patterns and insights from data obtained by observing complex and strange phenomena in the real world. Currently, data have become the key to economic growth and have been described as the “new oil.” In this climate, a new form of market for exchanging data with others for gaining new perspectives and interpreting and creating technologies are emerging. The data marketplace is where data and the products from data are treated as economic goods and the stakeholders of data businesses communicate with each other to cultivate a deeper understanding of the data usage. Even before the worldwide movements of big data and AI, authors have been focusing on information technology that contributes to interdisciplinary collaboration, knowledge sharing, and decision making, and have been conducting research. In this chapter, we show the fundamental concept and the meaning of the data marketplace for innovations with the role of data, the current state of the data businesses in the ecosystem, and their introduction examples from the social backgrounds and requirements.KeywordsMarket of dataData exchangeValue exchangeData platformInformationKnowledge
Article
Research into real-time simulation applications outside of manufacturing environments has extended to sociotechnical systems such as healthcare over the past decade, where a number of published studies have demonstrated proof-of-concept models for near-future resource planning. Using real-time decision-support systems, people take decisions supported by the output of simulations. However real-time simulation frameworks abstract human intervention to an “external decision-maker,” with little regard to the complexities of underlying decision-making constructs, and how design and development decisions can impact the quality of decision-support. One such construct is situation awareness (SA), which is a precursor to decision-making. It is a dynamic state of knowledge about how a situation is unfolding; one approach to enhancing situation awareness is the provision of appropriate real-time information. We argue that design, development and implementation decisions should be focused at the interface between decision-making and decision-support. This integrative literature review proposes a SA framework integrating models of SA with a technical perspective for real-time simulation, to support an understanding of the cognitive needs of users alongside technical details during the development process. The implications for the usefulness and usability of real-time decision-support tools are discussed with application to Emergency Departments.
Chapter
Various industries worldwide have been severely affected by the COVID-19 pandemic, highlighting the gaps between social systems and forcing major transformations of our lives. To understand and mitigate the phenomena related to the unprecedented danger of COVID-19, we have become acutely aware of the importance of data distribution, exchange, and sharing across fields; indeed, various data are published and used in decision-making processes. However, although many international organizations and companies have been publishing data and adopting relevant measures, data sharing regarding the question of what data are required for any purpose is insufficient; that is, data are principally provided by organizations who publish the data unilaterally; currently, data-related needs are not shared or leveraged. To address this issue, we introduce the concept of “data origination.” Data origination is the act of designing/acquiring/utilizing data that considers the subjective knowledge and diversity of perspectives of humans, and that aims to elucidate and support this process. We also discuss a case study of data needs and unexplored data externalization conducted during the COVID-19 pandemic, based on data origination.KeywordsData originationUnexplored dataData designData exchangeKnowledge sharing
Full-text available
Chapter
Zusammenfassung Dieser Beitrag stellt dar, welche Chancen und Herausforderungen mit der Bewertung von Daten sowie der Abbildung monetärer Datenwerte verbunden sind und geht auf mögliche Lösungsansätze zur Bewertung von Unternehmensdatenbeständen, insbesondere im Kontext der industriellen Produktion, ein. Zunächst werden Grundlagen zur Charakterisierung, Nutzung und Verwertung von Daten sowie bestehende Methoden zur Bewertung von immateriellen Vermögensgegenständen dargestellt. Darauf aufbauend werden Chancen und Herausforderungen spezifiziert, potenzielle Lösungsansätze zur Datenbewertung abgeleitet und anschließend Anforderungen für die Datenbewertung beschrieben sowie die nutzenorientierte Datenbewertung skizziert.
Article
It is argued that computing machines inevitably involve devices which perform logical functions that do not have a single-valued inverse. This logical irreversibility is associated with physical irreversibility and requires a minimal heat generation, per machine cycle, typically of the order of kT for each irreversible function. This dissipation serves the purpose of standardizing signals and making them independent of their exact logical history. Two simple, but representative, models of bistable devices are subjected to a more detailed analysis of switching kinetics to yield the relationship between speed and energy dissipation, and to estimate the effects of errors induced by thermal fluctuations.
Article
Many decisions are based on beliefs concerning the likelihood of uncertain events such as the outcome of an election, the guilt of a defendant, or the future value of the dollar. Occasionally, beliefs concerning uncertain events are expressed in numerical form as odds or subjective probabilities. In general, the heuristics are quite useful, but sometimes they lead to severe and systematic errors. The subjective assessment of probability resembles the subjective assessment of physical quantities such as distance or size. These judgments are all based on data of limited validity, which are processed according to heuristic rules. However, the reliance on this rule leads to systematic errors in the estimation of distance. This chapter describes three heuristics that are employed in making judgments under uncertainty. The first is representativeness, which is usually employed when people are asked to judge the probability that an object or event belongs to a class or event. The second is the availability of instances or scenarios, which is often employed when people are asked to assess the frequency of a class or the plausibility of a particular development, and the third is adjustment from an anchor, which is usually employed in numerical prediction when a relevant value is available.