ArticlePDF Available


In this paper we want to examine how the mutual understanding of speakers is reached during a conversation through collaborative processes, and what role is played by abductive inference (in the Peircean sense) in these processes. We do this by bringing together contributions coming from a variety of disciplines, such as logic, philosophy of language and psychology. When speakers are engaged in a conversation, they refer to a supposed common ground: every participant ascribes to the others some knowledge, belief, opinion etc. on which to rely in order to reach mutual understanding. As the conversation unfolds, this common ground is continually corrected and reshaped by the interchanges. An abductive reasoning takes place, in a collaborative setting, in order to build new possible theories about the common ground. In reconstructing this process through the use of a working example, we argue that the integration of a collaborative perspective within the Peircean theory of abduction can help to solve some of the drawbacks that the critics of the latter have outlined, for example its permissivity and non generativity.
Abductive Reasoning, Interpretation
and Collaborative Processes
Claudia Arrighi and Roberta Ferrario
1. Introduction
In this paper we are going to investigate the role of abductive reasoning in the interpretation
of natural language, arguing that the similarities go deeper than the ones shown in other works,
and that we can derive from this parallel some enrichment of the notion of abductive reasoning. In
talking about the interpretation of natural language, we refer specifically to the work of Donald
Davidson, which stresses the impressive human ability to reach mutual understanding in a
conversation, despite the numerous ambiguities, malapropisms, unusual expressions that, in
everyday communication, we put in almost every sentence. How is it possible that people who, at
the beginning of a conversation go from misunderstanding to misunderstanding, at the end of that
same conversation can reach an agreement of any sort?
Davidson's position is that this agreement cannot be reached by simply following the rules
of language, but we need extralinguistic abilities, namely the ability of forming provisional
hypotheses on the interpretation of the words uttered by other speakers, and the ability of revising
these hypotheses through the interaction with other speakers and the environment. This idea of
provisional hypotheses is at the core of the connection with abductive reasoning, as it is shown in
(Wirth 1999). However, we think that there is more to be said about this connection, in particular
regarding the role played by the interaction with other speakers and the environment. We consider
this aspect fundamental, because it can provide a framework to overcome the weakness imputed
to abductive reasoning. Hence, we propose a way to integrate this aspect in a schematization of
abductive reasoning, borrowing a structured way to represent interaction between speakers from
the work of Herbert Clark.
From a wider point of view, this work is meant to be a contribution to a current of thought
developing at the intersection of philosophy and computing, what has been called a form of
“socially sensitive computing” (Addis et al. 2004). Inside this approach, the contrast between
human cognitive abilities and formal models is mediated developing the idea of “irrational sets” as
product of abductive inferences, in various contexts of empirical and social interaction (Addis,
Billinge and Visscher 2004, Addis et al. 2005). As a consequence of some of our precedent work
on meaning (Arrighi and Ferrario 2004), we think that this idea of ‘not-well-defined’ sets is clearly
needed in the study of language and in natural language processing, as is shown, from a technical
point of view, in other in works, such as Vadera et al. (2005).
The paper is structured as follows: Section 2 presents the features of the problem of mutual
understanding, while in section 3 we sketch Davidson’s ideas on the process of interpretation. In
section 4 we recall the parallelism that Wirth traced between Davidson and Peirce, and section 5
gives a presentation of the collaborative and interactive aspects of language interpretation, what
we thought was missing in Wirth's paper. In section 6, based on Clark’s work, we try to embed
abductive inference in a collaborative framework; finally, section 7 contains the conclusions of the
2. The problem of mutual understanding
Long-lasting debates, in philosophy as well as in other disciplines, have shown that it is clearly
problematic to give a complete account of how we, human beings, communicate through language.
The difficulty is also evident when we look at all the concrete applications produced by researchers
in computer science and artificial intelligence in the field of Natural Language Processing, in order
to simulate human communication. Many of the issues are now well known and thoroughly studied,
nevertheless they still pose very serious obstacles to the technical mastery of human skills in using
language. Why is explaining how language works so problematic?
One of the main issues is that of ambiguity, in interpreting single words, in the structure of
sentences, and different uses of an expression. This ambiguity is manifest in everyday experience,
it is usually the source of most misunderstandings, but it is also the source of many jokes, as is
shown in this quotation from Groucho Marx:
Outside of a dog, a book is a man’s best friend.
Inside of a dog it’s too dark to read.
In this example, the joke results from two different - but in any case both ordinary - uses, of the
term “outside”. Things can get even more complicated with so-called non-literal interpretations,
when speakers make rhetorical use of language, as in metaphorical, ironical, or sarcastic
Another important source of problems is the role of background and contextual information:
very often, in conversations a lot of elements that seem fundamental for a correct understanding of
the message conveyed by a linguistic expression are left implicit, things like presuppositions,
implicatures, chunks of common knowledge etc. For example, if someone asks us “Do you know
the time?” it’s likely that she would like to know what time it is, and not just if we are informed about
it or not. Nothing in her sentence literally asks for the time, but we simply know that this is what she
intends. Let’s give another simple real life example. Roberta calls Claudia in her office and
Michael, her officemate, answers the phone:
Roberta: Hello, is Claudia there?
Michael: She is here, but not here
Roberta: Can I leave a message?
Michael: Shoot!
In this example, the answer “she is here but not here” seems to be informative for Roberta, instead
of being perceived just as an overt contradiction, as suggested by a logical analysis of the
As a consequence of considerations like these, the debate on language has often turned
around the question: Is the knowledge of words and rules of a certain language sufficient to explain
how we understand that language? And if so, how many rules do we need, can they all be spelled
out? One of the positions in the philosophy of language claims that speakers’ mastery of a
language cannot be represented by a set of words and rules of that same language. What is
required is also an extralinguistic competence, i.e. a variety of skills that are not necessarily
confined to the linguistic activity.
Amongst these extralinguistic skills which might be at the basis of language use, we are
particularly interested in the ability of building plausible theories on the intentions of the other
participants in the conversation. In other terms, each participant in the conversation forms some
hypotheses aimed at giving a correct interpretation of the utterances of the others and the
conversation is successful when such hypotheses fit sufficiently well what the interlocutors
intended. Therefore, there is no exact criterion able to determine the success of a conversational
interchange, only the participants can pragmatically guess if the information received was
sufficiently similar to the one that was intended to be conveyed. Donald Davidson is one of the
scholars who have suggested this view, and we want to spend few words in more details on his
3. Donald Davidson: interpreting as a process of formation and
testing of contextual hypotheses
One of the central questions addressed by Donald Davidson in his work on philosophy of language
is the so-called problem of radical interpretation, i.e. what makes it possible for two speakers of the
same language to share the meaning of a certain word or utterance? How do we know that it is the
same meaning? Initially raised by Quine as a problem in dealing with different languages (as the
problem of radical translation), it was soon extended to the broader question of how we know that
some term means the same thing for two, or more, individuals.
Davidson absolutely rejects the common sensical explanation that we share language,
arguing that the notion itself of “sharing a language” could not be defined inside the same
“[we] must give up the idea of a clearly defined shared structure which
language users acquire and then apply to cases” (Davidson 1986).
He argues instead for the necessity of a relation between linguistic meaning and extralinguistic
context of an utterance, he strongly advocates a “contextualized” interpretation of language by
means that are not part of the language itself. The importance of these extralinguistic means is
particularly evident in observing what makes possible for the speakers to correctly react to
ambiguous, unconventional or non codified uses of language and, at the same time, enables a
creative or distorted use of language, as in metaphors or malapropisms.
In (Davidson 1986), Davidson gives an interesting example based on malapropisms in
order to explain how understanding works in conversation. Davidson moves from the notion of first
meaning, namely the first intended outcome of a speaker’s utterance. The fact that often, in
presence of a speaker who uses improperly some terms, a hearer appears nonetheless to grasp
the first meaning comes as a surprise if we accept the theory that speaker and hearer share a
common language where words mean – literally – something that as been already fixed and
In the solution proposed by Davidson, the hearer comes to the conversation with a prior
theory of how utterances should be interpreted; but during the conversation, if something doesn’t fit
this theory, she builds a passing theory to accommodate the discrepancy. This latter theory is built
ad hoc to interpret this precise speaker in this precise conversation and can then be either
abandoned or re-used in other conversations with the same speaker or with different interlocutors.
The ability to build a passing theory is anchored in human skills that enable us to retrieve
an interpretation that makes sense of the speaker’s utterance (in the case of Davidson’s article, a
similar sounding word has to be substituted to the one actually uttered by the speaker). These
skills are something that we acquire and cannot be systematically described or taught.
4. Uwe Wirth: parallelism between Davidson’s idea and Peirce’s
abductive inference
In his interesting paper (Wirth 1999), Uwe Wirth has drawn a parallel between Davidson’s ideas on
interpretation as formulation of theories and the process of abductive inference presented by C.S.
Peirce in many of his works.
Abduction is one of Peirce's four steps of scientific investigation; it has been called a "logic
of discovery”, even though it is not an argument for validating scientific hypotheses. These four
steps are anomaly observation, abduction of hypothesis, inductive testing, and deductive
confirmation. The idea is that, after having observed an unpredicted phenomenon, a reasoner
starts to look for an explanation; this explanation is reached by abduction and is only tentative.
Based on this explanation, the reasoner - through a deduction process - forms a new prediction
and then, finally, performs some inductive test in order to verify the prediction.
Peirce writes:
Abduction is the process of forming an explanatory hypothesis. It is the only
logical operation which introduces any new idea; for induction does nothing but
determine a value, and deduction merely evolves the necessary consequences
of a pure hypothesis.
Deduction proves that something must be; Induction shows that something
actually is operative; Abduction merely suggests that something may be.
(Peirce, CP 5.181)
The abductive inference is described by Peirce himself in the following way:
[1] The surprising fact, C, is observed;
[2] But if A were true, C would be a matter of course.
[3] There is reason to suspect that A is true
(Peirce, CP 5.189)
Where C is the surprising fact and A a set of provisional hypotheses.
Peirce’s “surprising fact” is the occurrence of an unexpected event, which is inconsistent
with the assumptions and thus generates surprise and calls for an “accommodation” of the theory.
In other words, the inconsistency yields the search for some unknown premises that should have
been taken into account but weren’t. Once that these new premises have been found (abduced),
they change the context in which the observation of the surprising fact has taken place, therefore
eliminating the inconsistency.
As explained by Wirth, something very similar happens in Davidson’s views on
conversational activities. When two agents A and B are engaged in a conversation, they make
some tacit assumptions about each other and about how a communicative exchange should work.
These tacit assumptions include, for example, Grice's well known imperatives, such as expecting
the speaker to tell the truth, to be as informative as required, to respond in a way that is relevant to
the topic of discussion. Obviously, it is not always the case that these requirements are met; if a
sentence uttered by A violates one of these assumptions, the violation qualifies as “surprising
event” for B (in Peircean terms). Instead of rejecting the sentence, B performs an abduction of a
premise that could make sense of A’s contribution, thus enriching her interpretation of a new,
previously neglected, element. We can give an example of this going back to the conversation on
the phone, in order to analyze it under the new perspective we have acquired thanks to the
suggestions given by Wirth’s paper. In that conversation, the sentence under focus is the following:
Michael: She is here, but not here
If we paraphrase Peirce, we have:
[1] Michael’s sentence is a contradiction, i.e. a surprising fact, because it
violates common principles of communication;
[2] But if “here” referred to two different things (the room and the building),
Michael’s sentence would be informative.
[3] There is reason to suspect that “here” refers to two different things.
Instead of being stuck in the apparent contradiction, Roberta performs an abductive
process of reasoning, from which she concludes that Michael intended to communicate that
Claudia is not in the same room where he is, but she is in the building, and this is informative for
Roberta. From this conclusion, Roberta goes on in the conversation with this new
assumption/interpretation about Michael's intended message, so she now thinks that she can leave
a message for Claudia to Michael, which will be delivered to her as soon as she comes back in the
room. Here is where Wirth finds the analogy between the two approaches: generally speaking,
both these processes are seen as the “provisional adoption of an explanatory hypothesis” (Peirce,
CP 4.451), when an inconsistency is found in the theory at hand. Even if Davidson is addressing
his analysis to language, he argues that in order to correctly interpret a language, competent
speakers make use of a skill that is pre-linguistic, namely the ability to create a plausible theory
about the speaker’s intention.
Wirth suggests another common point between the two theories, the fact that, in his
opinion, in both approaches the abductive competence is obtained by the interplay of an “economy
of discourse” with the well known principle of charity, introduced by Neil Wilson (Wilson 1959, p.
532) and advocated by Davidson in interpreting language. The principle of charity roughly says that
when we interpret a sentence in an unintelligible way, it is more likely that our interpretation is a
bad one, rather than a good interpretation of an unintelligible view. Thus the principle encourages
to revise the interpretation in case of a mismatch with the expectations and the revision should be
directed to the maximization of the economy; in other words the solution that requires the minimum
cognitive, physical and economic effort should be taken.
This principle is certainly present in Davidson’s works and is useful in order to trace the
similitude with Peirce’s abduction, but it shows also that the two approaches share the same
weakness: we have no clear indication of where to find alternative, charitable, interpretations, or
what should be the parameter to evaluate the 'cost', in terms of cognitive effort, of each of these
alternative options. A theory of language interpretation relying on such a fuzzy principle can be
criticized for being too vague, because it does not indicate where to look to find a measure of what
is charitable or economically advantageous. This criticism seems to have much in common with
the criticisms traditionally moved against abduction as an instrument to give an account of
reasoning and discovery. As well explained in (Paavola 2004), abduction has been criticized under
two perspectives: on the one hand, it is seen as too permissive, since an infinite number of
hypotheses can be generated in every process of abduction and this is because it doesn’t provide
a criterion of relevance for such hypotheses. On the other hand, it is not generative since the
missing hypothesis we are looking for (A) is already present in the premise 2 (But if A were true, C
would be a matter of course), i.e. it is supposed to be known in advance, as opposed to be
generated by the abductive process.
The paper by Wirth has definitely the merit of pointing out some of the main analogies in
these particular aspects of the works of Davidson and Peirce, but unfortunately does not provide
articulated suggestions on how to overcome this weakness. This is exactly what we would like to
do, underlining the intersubjective nature of language and the importance of interaction with the
environment as something that provides a framework of reference in formulating hypotheses as
presented in the next section and then look again at the abductive inference, to see if something
similar is applicable.
5. The collaborative and interactive aspect, a fundamental feature for
language interpretation
We want to argue that in Wirth’s parallelism something has been left out, that could be a promising
source of constraints to the charitable efforts in interpreting, or, on the other side of the parallelism,
a source of missing premises to be tested in an abduction process. There is something in
Davidson’s ideas on interpretation that can be integrated with abductive reasoning, and which
Wirth does not point out sharply enough, namely the importance of interaction with other speakers
and with the entities in the environment, beyond the mere analysis of the linguistic expression.
When dealing with natural language and acquisition of words, abductive reasoning takes
place in an interactive setting, situated in a precise environment, or empirical context, where an
essential part of this context is constituted by the speaker’s and hearer’s viewpoints, and another
fundamental part is given by the fact that speaker and hearer share a world.
In (Davidson 1982) Davidson says:
To understand the speech of another, I must be able to think of the same
things she does; I must share her world.
[…] Communication depends on each communicator having, and correctly
thinking that the other has, the concept of a shared world, an intersubjective
(Davidson 1982)
Elsewhere, Davidson uses the metaphor of the triangle to show how the formation of a concept (for
instance of a table or a bell) has its roots in the interplay of (at least) two human beings and a
shared world:
“It takes two points of view to give a location to the cause of a thought, and
thus to define its content. We may think of it as a form of triangulation: each of
two people is reacting differentially to sensory stimuli streaming in from a certain
direction. [...] If two people now note each other’s reaction (in the case of
language, verbal reactions), each can correlate these observed reactions with his
or her stimuli from the world. A common cause has been determined. The
triangle which gives content to thought and speech is complete. But it takes two
to triangulate"
(Davidson 1991)
[…] to have the concept of a table or a bell is to recognize the existence of a
triangle, one apex of which is oneself, the second apex another creature similar
to oneself, and the third an object (table or bell) located in a space thus made
(Davidson 1992)
This interaction must be conceived as something much more complex than just reception of
stimuli. As argued by Dagfinn Føllesdal in (Føllesdal 1995) talking about the theory of meaning in
Davidson, language has to be rooted on the “perception of publicly observable physical objects
and events”, something that is subject to social assessment. Our response to a certain stimulus is
due also to the anticipations and expectations surrounding the reception of such stimulus:
“we group the stimuli and responses in equivalence classes in view of our
tentative theory of what the subject means and does”
(Føllesdal 1982).
These quotations should give an idea of how central, in Davidson's conception of language,
the interactive and situated characters are. In our opinion, following the lines of the parallelism
traced by Wirth, these aspects of language are very relevant also with respect to the discussion on
the weakness of abduction. They could represent the missing ingredients to give a fuller account of
abductive reasoning, providing a relevance criterion for hypotheses selection, thus showing that
they are not already known ab initio by the reasoning subject, but they are the product of an
interactive process, both with the environment and the others speakers.
In the case of language, different individuals enter a conversation with different prior
theories about the subject matter and the opinions of the other participants. These prior theories
shape expectations about how the conversation will proceed; every time one of the expectations of
an individual is not met (i.e. a surprising event happens), this individual is naturally brought to
revise her prior theory and to adopt a new passing theory (that includes the new hypothesis that
accommodates the surprising fact). The new hypothesis is not created or selected at random, but it
strongly depends on what was going on in the previous part of the conversation, or in the activity
shared with the others speakers; the goal of the activity, the reactions of the other participants
and/or the features and inputs of the shared environment are what shapes the passing theory of
the speaker, what determines the adoption of a certain hypothesis among all possible ones.
Can we apply a similar idea to abductive reasoning? The application of this framework to
the study of abduction should not sound as completely new, since other analyses of abduction
have already suggested that the usefulness of this tool in reasoning can be fully appreciated only if
embedded in a framework of interaction, either with an environment or with other agents. For
example, in a recent collection of papers on abductive reasoning in science, Magnani (Magnani
2004) has presented a manipulative framework, according to which concrete manipulations of
external objects influence the generation of hypotheses” (p. 243), and Paavola (Paavola 2004) has
argued in favor of a strategic framework, that further develops an idea found in Hintikka:
“the validity of an abductive inference is to be judged by strategic principles
rather than by definitory rules”
(Hintikka 1998, p. 513)
“strategies are related to goal-directed activity, where the ability to anticipate
things, and to assess or choose between different possibilities, are important”
(Paavola 2004, p. 270).
These views summarize well the two aspects that we have underlined as fundamental. In
formulating theories, new hypotheses are not floating in a vacuum, waiting to be picked up, but
they can be suggested by our interaction with the present environment and the other subjects
involved; and because this interaction is most of the times goal-oriented, the selection of the
hypotheses worth to be tested is guided by strategic considerations, with respect to the goal to
reach. We find very significant that in other works Paavola and colleagues have developed, along
with the approach to abduction based on strategy, also the idea of a “trialogue”, very much in the
spirit of Davidsonian idea of triangulation (Paavola, Hakkarainen and Sintonen 2005).
To explain better these ideas, we would like to show in a more concrete fashion how an
abductive inference can be embedded in an interactive and goal-oriented process, and what kinds
of elements provide constraints to the number of provisional hypotheses that can be adopted. In
order to do this, in the next section we introduce a basic structure of a linguistic exchange as
represented by the psychologist Herbert Clark.
6. Clark’s rendition of the collaborative framework of mutual
The psychologist Herbert Clark has provided an interesting formulation of the interactive and
strategic character of language in his book Using language (Clark and Wilkes-Gibbs 1996). He
describes linguistic communication as a joint activity, i.e. an activity carried out by a group of
people acting in coordination with each other, in order to reach a common goal, that is the mutual
understanding of each other’s utterances (Clark and Wilkes-Gibbs 1986). Usually this joint activity
is embedded in other joint activities of higher level, for example the conversation we are having is
in the context of preparing dinner together - building a hierarchy of different levels of activity.
Coordination and the goal-oriented nature are very much central points in Clark’s representation of
linguistic exchanges, and he goes deep in trying to identify as many elements as possible which
the coordinating process can rely on. Amongst these elements we would like to pick out two that
we consider particularly important, the common ground and the common construal:
- Common ground: the information that the speakers have in common
- Common construal: a model of the evolving present activity, including the
identification of elements that are salient for this activity, mutual expectations on the
roles played by every speaker and on the moves needed to reach the goal (mutual
The aim of the participants in a conversation (aside form the individual goals that led to the
conversation itself) is to find an agreement on the meaning of what is said. In a very crude
summarizing effort, we could say that mutual understanding is a process of verification of the
common ground and of establishment of a common construal. This goal can be reached as it is in
many other joint activities: by acting together according to some intended principles, using salient
objects or features in the environment and choosing our next move depending on which moves
have already been made. Clark describes these steps in more details in his co-authored paper
“Referring as a collaborative process” (Clark and Wilkes-Gibbs 1996, pp.20-24):
- The steps: Presentation, Evaluation, Rejection, Refashion, Acceptance, Follow-ups.
The best way to describe these steps is to give an example. The following is a slightly
modified version of an example given in (Arrighi 2004), where the steps are used in a joint effort to
verify a discrepancy in the assumed common ground, and to reach a common construal about the
meaning of the word ‘bread':
Sonal: “Can you bring the bread to the table,
Sonal makes the presentation
Viola: “Bread? There isn’t any bread”
Viola evaluates, then rejects. Sonal’s move is
not enough to reach the goal, to make herself
Sonal gives Viola a basket with naan bread.
Sonal refashions her presentation, with an
ostensive act to make an object salient
Viola: “Oh, is this bread?”
Viola partially accepts, but asking for
confirmation, a follow-up
Sonal: “Yes, it is naan bread”
Sonal refashions her ostensive act as an
Viola: “I see... I didn’t know”
Viola accepts
If we frame communication as a game, we can think of the speakers as players, guided by
the goal of mutual understanding and constrained by many rules (the common ground), some
about language, some about social interaction, some derived by physical constraints - exactly
which rules apply depends on the stage of the game. In order to reach the goal, the players need
to formulate a clear representation of the actual situation that is on the game-board (the common
construal), and accordingly they think about a winning strategy, changing it as needed, depending
on the moves of the other player.
The strategy is formulated with the final goal of the game in mind, and the elements of the
context relevant to the game (for instance, an open door is not relevant to formulate a winning
strategy in a chess game). When the other player makes a move that blocks our winning strategy,
abductive reasoning enters the picture: which other possibilities do I have to rebuild a winning
The way in which Clark presents the process of communication as a joint activity aimed at
constructing a common ground for meaning seems to be a very accurate and plausible description
of how things happen in linguistic communication. Now we would like to use Clark’s schematization
to compare the steps for reaching mutual understanding with the steps in the reasoning processes
that characterize abduction.
7. Embedding abductive inference in a collaborative framework
Let’s go back to the dialogue between Viola and Sonal, this time formulating the steps as an
abductive inference. In this case we have a collaborative game with a common goal, the two
players are supposed to reach mutual understanding but nevertheless the moves of one player
can be an obstacle for the other in some circumstances.
Sonal: “Can you bring the bread to the
table, please?”
Viola`s assumptions include: Sonal is referring to
something, the reference should be salient and
Viola: “Bread? There isn’t any bread”
Surprising fact: the reference is not available.
Viola questions Sonal about this incongruence
Sonal gives Viola a basket with naan
Sonal performs an action, making an object
salient. Given the general goal of their exchange,
Viola has the assumption that Sonal’s action
must be relevant to the main goal
Viola: “Oh, is this bread?”
If this object was the intended reference, her
request would be legitimate: Viola tests the new
Sonal: “Yes, it is naan bread”
Sonal confirms the new common construal
In this example Viola is the player who needs to use abduction in order to reach the goal of
interpreting Sonal’s utterance. Given her assumptions about their common ground (for instance,
they both speak English), and about the common construal (including that they are preparing
dinner together), Viola understands that Sonal’s utterance is a request, and if Viola wants to fulfill
this request, she has to pick the reference of the word ‘bread’ and put it on the reference of the
word ‘table’. It is also assumed that Sonal wants to be understood by Viola, and therefore the
references of the words must be available in the environment.
Here comes the surprising fact: there is no reference for the word ‘bread’. It is surprising
because it is not consistent with the common ground or with the common construal that Viola
thinks to be sharing with Sonal about how a conversation should work. But the interaction does not
fail just because an inconsistency has been found – having a winning strategy in the game
requires that difficulties must be overcome by making the proper move. In the present situation, the
proper move is the simplest, acknowledging the problem and waiting for help: “Bread? There isn’t
any bread”.
After the surprising fact, Viola does not stop formulating hypotheses on her own, but she
chooses to interact with the environment, in this case to interact with Sonal. When Sonal responds
with an action to Viola’s utterance, Viola assumes that Sonal’s action must be relevant to solve the
“surprising fact” - the object shown must be the object Sonal intended. Viola, involved in a
collaborative game, goal oriented and focusing on an effective winning strategy, does not stop
considering all the plausible hypotheses, but just the one suggested in the present situation,
immediately testing it: “Oh, is this bread?”
How does the join activity framework affect the selection of hypotheses? After Sonal’s
request, is it possible that, looking around the kitchen, Viola sees a still nature on the wall featuring
a loaf of bread. Can this have been taken as a plausible reference? Well, in many contexts it could,
but in the present situation Viola judges this option not relevant for the common goal, for the joint
activity in which this conversation is embedded. Viola is testing her assumption that Sonal is
referring to a salient object in the room referred to by the word ‘bread’, but it must be salient in the
context of the higher level assumption that they are engaged in the join activity of getting ready for
a dinner. With her question, Viola expresses her difficulty with the assumption of lower level, but
she still keeps the higher level assumption. This could be envisioned as a measure of the cognitive
effort requested in revising the theory.
8. Conclusions
In this paper we have explored the common traits in interpreting language and in abductive
reasoning. We started by considering a comparison between Davidson and Peirce put forward by
Uwe Wirth, where an abduction process is compared to Davidson’s idea that linguistic
interpretation relies on the speakers’ ability of building tentative hypotheses about the context and
the intentions of the other speakers. These hypotheses can be revised when a new event does not
fit the model built so far - but how do we choose new hypotheses?
The traditional weaknesses attributed to abduction have been summarized as the faults of
being too permissive and non generative, and these faults seem to be the counterpart of the
objections moved against the fuzziness of Davidson’s ideas of the principle of charity and
economy, in formulating new interpretations. Can these difficulties be overcome? We thought that
we should look closer to some fundamental features of Davidson’s account of interpretation, in
particular, the interaction with both the environment and other speakers seems to be fundamental.
Using a structured account of this interaction, as provided by the psychologist Herbert Clark, we
have analyzed whether this ‘structure’ can be of any help in generating and selecting alternative
The first problem was permissivity: abduction doesn’t provide a criterion to choose which
assumptions were disregarded when formulating a failed explanation of a phenomenon. The
“surprising event” tells us just that something is missing, but not what is missing, so any
assumption can be taken in order to accommodate the anomaly. But, as happens in
communication, the agent who performs abduction doesn’t face the surprising event alone, she is
embedded in an environment and in a social context that react to her attempts of accommodation,
thus she can negotiate with the environment and with the other agents what was missing from her
previous hypotheses. In other terms, the limits of the shared environment and social context,
conceived as a common ground and a common construal, are a constraint for the range of
hypotheses to evaluate.
The interaction with environment and social context also offers the solution to the other
problem, namely non generativity. The problem emerges from the fact that in the classical
formulation of abduction, the new hypothesis that should accommodate the surprising event seems
to be forcedly known in advance by the subject who performs the abduction. But, again, this is
consequence of the fact that we used to see the subject in isolation, while in real situations
environment or the social context can offer new hypotheses, previously unknown to the reasoner.
Moreover, the goal and the rules of the activity in which the reasoning is embedded drive in a
certain direction the research for new hypotheses.
The final claim of the paper is then that an accurate account of the success in
communication must be based on an integration of abductive with cooperative processes, and this
is a direction of research definitely worthy of further exploration.
Addis, T., Billinge, D., Visscher, B. (2004). Socially Sensitive Computing: a Necessary Paradigm
Shift for Computer Science. Grand Challenges in Computing, UK CPHC, Newcastle, 2004,
pp. 1-6.
Addis, T., Townsend-Addis, J., Billinge, D., Gooding, D. C., Visscher B. (2005). Tracking Irrational
Sets. MBR 2004, Pavia, Italy.
Arrighi, C., Ferrario, R. (2004). The Dynamic Nature of Meaning. ECAP 2004, Pavia, Italy. [year
even if unpublished?]
Arrighi, C. (2004). Meaning as a Product of Negotiation. Master's Thesis, Philosophy Department,
Stanford University.
Clark, H. H., Wilkes-Gibbs, D. (1986). Referring as a collaborative process. Cognition 22: 1-39.
Clark, H. H. (1996). Using Language. New York: Cambridge University Press.
Davidson, D. (1982). Rational Animals. Dialectica 36: 317-327.
Davidson, D. (1986). A Nice Derangement of Epitaphs. In R. Grandy and R. Warner (eds.)
Philosophical Grounds of Rationality. Oxford University Press, pp. 433-446.
Davidson, D. (1991). Three Varieties of Knowledge. In Subjective, Intersubjective, Objective.
Oxford: Clarendon Press, pp. 205-220.
Davidson, D. (1992). The Second Person. In Subjective, Intersubjective, Objective. Oxford:
Clarendon Press, pp. 107-122.
Føllesdal, D. (1982). Intentionality and Behaviorism. In L. J. Cohen, J. Los, H. Pfeiffer, and Klaus-
Peter Podewski (eds.) Proocedings of the 6th International Congress of Logic, Methodology
and Philosophy of Science, Hannover August 22-29, 1979. Amsterdam: North Holland.
Føllesdal, D. (1995). In What Sense Is Language Public? In P. Leonardi (Ed.), On Quine: New
Essays. New York & Cambridge: Cambridge University Press.
Grice, H. P. (1969). Utterer's meaning and intentions. The Philosophical Review, 78: 147-177.
Hartshorne, C., Weiss P. (eds.) (1931-1958). Peirce, Charles Sanders. Collected Papers.
Cambridge MA: Harvard University Press.
Hintikka, J. (1976). Language-Games. Acta Philosophica Finnica, XXVIII(1-3): 105-125.
Hintikka, J. (1998). What is Abduction? The Fundamental Problem of Contemporary Epistemology.
Transactions of the Charles S. Peirce Society, 34(3): 503-533.
Lewis, D. (1969). Convention: A Philosophical Study. Cambridge MA: Harvard University Press.
Magnani, L. (2004). Model-Based and Manipulative Abduction in Science. Foundations of Science,
9: 219-247.
Oltramari, A., Borgo S., et al. (2003). Negoziati di significato. Sistemi Intelligenti, 2003(3):419-428.
Paavola, S. (2004). Abduction as a Logic and Methodology of Discovery: the Importance of
Strategy. Foundations of Science, 9(3): 267-283.
Paavola, S., Hakkarainen, K., Sintonen, M. (2005) Abduction with Dialogical and Trialogical
Means, MBR 2004 Pavia, Italy.
Rapaport, W. J. (2003). What Did You Mean by That? Misunderstanding, Negotiation, and
Syntactic Semantics. Minds and Machines, 13(3): 397-427.
Vadera, S., Rodriguez A., Succar E., Wu J. (2005). Using Wittgenstein’s Family Resemblance
Principle to Learn Exemplars. MBR 2004, Pavia, Italy.
Wilson, N. (1959). Substances without substrata. Review of Metaphysics, 12(4): 521-539.
Wirth, U. (1999). Abductive Reasoning in Peirce's and Davidson's Account of Interpretation.
Transactions of the Charles S. Peirce Society, 35(1): 115-128.
Wittgenstein, L. (1958). Philosophical Investigations. New York: The Macmillan Company.
... Abductive logic is a tool for discerning a pattern out of the mystery (Martin, 2007b(Martin, , 2009. After an observation of an unpredicted phenomenon is made, abduction is used to find answers because it is perfect for managing incomplete information (Arrighi & Ferrario, 2008;Hintikka, 1998). In addition, an important feature of generative reasoning is also a trial-and-error concept, which is shown to foster innovation (Cannon & Edmondson, 2005;Thomke, 2003). ...
... NPD teams embracing the principle of collaborative abduction go beyond the conventional goals of cross-functional collaboration enabled by improved project planning and execution (Kahn, Barczak, and Moss, 2006;Kahn et al, 2012;Wheelwright and Clark, 1992). Through collaborative abduction, NPD teams engage in expansive and contextual learning (Boland and Collopy, 2004;Edmondson and Nembhard, 2009), resulting in novel concepts that cannot be produced solely by induction (inducing the problem through customer observations) or deduction (deducing solutions from how customers react to current product features; Arrighi and Ferrario, 2008). Novel ideas emerge when the team collates diverse ideas from a broad network of stakeholders and synthesizes them by blending analysis and intuition, making abductive leaps to imagine alternative solutions and worlds (Micheli et al., 2018). ...
Full-text available
Team design thinking in new product development (NPD) refers to a set of design‐oriented principles that can be embedded in the everyday routines of NPD teams. We theorize that team design thinking enables NPD teams to develop more innovative products–products that are useful and novel–by reducing the routine and cognition‐based inertia that constrains the team’s innovativeness. Through a study of 247 NPD projects, we investigate the relationships between team design thinking, product utility, and product novelty. Further, we also explore whether the team’s unfamiliarity with product contexts influences these relationships. Findings show that team design thinking is positively associated with useful products in both familiar and unfamiliar contexts, but that team design thinking is associated with product novelty only in familiar contexts. Findings also suggest that team design thinking can be framed as a dynamic capability that addresses routine and cognition‐based inertia by expanding the organization’s knowledge resource base, particularly, in familiar contexts. A key implication of this research is that team design thinking will influence product utility in a variety of situations, but its influence on product novelty will be evident primarily in familiar contexts.
... In this paper I have contended that the concept of manipulative abduction is of 938 particular interest in delineating the actual practice of abduction but also in the 939 development of computationally adequate programs used to rediscover, or discover 940 for the first time, for example, scientific hypotheses and mathematical theorems or 941 laws but also to favor the design of more realistic and efficient robots. 942 23 Cf. also Arrighi and Ferrario's (2008) study on mutual understanding that emphasizes the role in abductive reasoning of the collaborative processes involved in interaction with other speakers and with the entities of the environment; they explicitly refer to my approach called "manipulative framework" and to the "strategic framework" described by Paavola and his collaborators. The fact that science is inherently seen as a social process is emphasized by Addis and Gooding (2008), who argue that abduction does not work in isolation from other inference mechanisms and use game theory to relate the abductive system to actions that produce new information. ...
This article aims at introducing the concept of manipulative abduction, by showing how we can find at various levels of cognitive activity various methods of manipulative constructivity. From this perspective, also creativity and discovery are no longer seen as mysterious irrational processes, but, thanks to constructive accounts, they are viewed as complex relationships among different inferential steps that can be clearly analyzed and identified. I maintain that the analysis of the blending of external and internal representations and of what I have called epistemic mediators is important not only to delineate the actual practice of manipulative abduction but also to further enhance the development of programs computationally adequate in rediscovering or discovering for the first time, for example, scientific hypotheses or mathematical theorems and to favor the design of more realistic and efficient robots. I will focus on those particular kinds of abductive cognition that resort to the existence of extra-theoretical ways of thinking – thinking through doing – centering the attention to the fact that many cognitive processes are centered on “external representations,” as a means to create communicable accounts of new experiences ready to be integrated into previously existing systems of experimental and theoretical practices.
... Hence, Nyrup, despite being supportive of the legitimacy of abduction as a singular inference type, emphasizes that abduction is mainly relevant for scientific research if it is followed up by induction (Nyrup 2015, p. 750). This is in line with a third tendency, which views abduction as only one element of a research process (see fx Margolis 2007;van Benthem 2007) where the quality of abduction is mainly judged as in proportion to the actual corroborations of the hypothesis generated (Arrighi and Ferrario 2008). ...
Full-text available
In this article, we investigate the notion of abduction, as it relates to and can be applied in scientific research. We investigate abduction through philosophy of science and the purpose is to survey some of the way abduction can be used in scientific research to the same extent as in- and deduction can be used. First, we perform a walkthrough of an illustrative sample of literature, which suggests that abduction is an ambiguously used concept, and that abduction is often defined much narrower than in- or deduction. We then proceed to give an outline of the most vital treatments of abduction in modern times, and we try to define various modalities of abduction, both as an autonomous research strategy and inference type, but also in relation and contrast to in- and deduction. The article ends with some perspectives on philosophy of science in relation to abduction, and some suggestions for applications of abduction in the different sciences.
... They also use techniques such as mind mapping and brainstorming to generate and evaluate new ideas (Goel & Singh, 1998;Sutton & Hargadon, 1996). Open minded discourse generates cognitive conflicts which trigger abductive reasoning resulting in new concepts that inductive (learning from customers) or deductive reasoning (deriving product features based on current 'espoused theory' of customer behavior) may not produce (Arrighi & Ferrario, 2008). ...
Full-text available
Through a survey of 247 digital product development leaders, we explore how design thinking principles embedded into product development routines influence successful market outcomes. We find that product utility partially mediates the relationship between embedded design thinking principles and market success, and that this mediated relationship is stronger when the product team is confronted with an unfamiliar design problem. Surprisingly, and in contrast, product novelty does not mediate the relationship between embedded design thinking principles and market success. Instead, product novelty has a non-linear (inverted U) relationship with market success, but the application of design thinking principles flattens this curve and thus improves the likelihood that product novelty may lead to market success. Our findings, to an extent, explain the mixed results reported by prior studies examining the effects of design thinking on product and market outcomes.
... According to Paavola and Hakkarainen (2005) and Arrighi and Ferrario (2008), abductive inference is a "weak" form of inference. It selects and uses clue-like signs to proceed toward tentative hypotheses. ...
In design and learning studies, an increasing amount of attention has been paid to experience. Many design approaches relate experience to embodiment and phenomenology. The growth in the number of applications that use the Internet of Things (IoT) has shifted human interactions from mobile devices and computers to tangible, material things. In education, the pressure to learn and update skills and knowledge, especially in work environments, has underlined the challenge of understanding how workers learn from reflection while working. These directions have been fuelled by research findings in the neurosciences, embodied cognition, the extended phenomenological–cognitive system and the role of emotions in decision-making and meaning making. The perspective on experience in different disciplines varies, and the aim is often to categorise experience. These approaches provide a worthwhile view of the importance of experience in learning and design, such as the recent emphasis on conceptual and epistemological knowledge creation. In pragmatism, experience plays a considerable role in research, art, communication and reflection. Therefore, I rely on Peirce’s communicative theory of signs and Dewey’s philosophy of experience to examine how experience is connected to reflection and therefore how it is necessarily tangible.
Here, I focus on deviations of intent, from that expressed by the standard or ordinary use of language, in instances where abductive reasoning plays a necessary role. In such cases, speakers usually utter some “mysterious” words based on the assumed abductive capability of their interlocutors. In order to cause the desired effect in the audience, the speaker relies on the hearer’s knowledge of the relevant common dialogue framework and thus uses the utterance as an abductive trigger. For purposes of explanation, I present a curious kind of (pseudo)fallacy: ironic (pseudo)fallacy.
Full-text available
Supporting the development of creative competency is important for the actual challenges of the society. However, creativity has been mainly approached in an individual way, without considering the specificities of team-based creativity processes. In this chapter, we establish the differences between creativity as an individual approach and creativity as a collaborative process. Then we discuss creativity from the perspective of the leaners’ and teachers’ attitudes. Subsequently, we discuss the concept of the margin of creativity in different learning activities. We finalize this chapter by discussing digital uses that can support creativity in team-based contexts.
Conference Paper
Full-text available
Data visualization has been widely utilized in various scenarios in data analytics for business purposes, especially helping novice readers make sense of complex dataset with interactive functions. However, due to the insufficient theoretical un-derstanding of the process of developing interactive functions and visual presen-tations, interactive data visualization tools often display all available automatic graphing functions in front of users, instead of guiding them to choose a visuali-zation based on their demands. Thus, this paper is intended to construct a process of developing interactive visualization with a specific focus on enabling the in-teroperation between design and interpretation. Stemmed from organizational se-miotics, an abductive process will be portrayed in this paper to interpret the pro-cess of developing interactive data visualization. Especially the interactive func-tions will be employed in an iterative process, where producers can be aware of and respond to readers’ information demands on semantic, pragmatic and social levels.
Full-text available
Sommario L'ontologia – "lo studio dell'essere in quanto essere", scriveva Aristotele – nasce ed è usualmente concepita come una disciplina strettamente filosofica, lontana dal mondo della tecnologia. Negli ultimi anni, tuttavia, l'esplosione delle comunicazioni in rete ha favorito un nuovo fenomeno: gli aspetti ontologici dell'informazione hanno acquistato un crescente valore strategico. Mediante una caratterizzazione ontologica dell'informazione, infatti, questa può essere reperita, isolata, organizzata e integrata in base a ciò che più conta: il suo contenuto. Questo breve articolo si propone di presentare in modo informale un aspetto fondamentale delle ricerche in questo campo, legato al problema della disambiguazione del significato. Dopo l'ultima esplorazione Davide aveva contratto una violenta forma influenzale, o almeno così credeva, dal momento che aveva letto di sintomi del genere solo nel manuale di archeomedicina, ai tempi dell'università. Febbre alta, mucosa irritata, difficoltà respiratorie, cefalea: la situazione peggiorava d'ora in ora, mettendo a repentaglio la riuscita della missione. «Ci deve essere un problema con i filtri per il trattamento dell'aria» pensò Davide, responsabile dell'ultima missione spaziale programmata per il 2103. Davide decise così di ricorrere a JIMMIE, sistema di intelligenza artificiale con cui erano equipaggiate tutte le astronavi dell'epoca. JIMMIE poteva elaborare dati di vario genere e interagire con utenti umani mediante un classico procedimento di domanda/risposta. Caricato il referto dei test clinici e inserita la query 'cura', Davide aveva visto comparire sullo schermo olografico le specifiche del nanobot 1 da impiegare contro il virus influenzale: purtroppo nulla di così peculiare era disponibile sul vascello, provvisto solo di dispositivi standard (come accadeva regolarmente ormai su tutte le unità della flotta, a causa dei tagli effettuati dall'Agenzia Spaziale) e costruire un nanobot ex-novo avrebbe richiesto tempi troppo lunghi. JIMMIE non sembrava possedere in memoria alcuna traccia né dei trattamenti anti-influenzali tipici del Ventesimo Secolo, né delle medicine che venivano usate a quell'epoca. Davide era sicuro dell'efficacia di tali trattamenti (che però richiedevano tempi abbastanza lunghi, diversamente dai nanobot che avevano un effetto quasi istantaneo), ma purtroppo non ne aveva che un vago ricordo accademico. 1 Recente generazione di robot multifunzionale miniaturizzato, frutto di ricerche avanzate in nanoscienza, che viene iniettato nel corpo umano per risolvere anomalie di vario tipo.
“Behaviorism” stands for a variety of attitudes and methodological positions in psychology. There are all kinds of positions whose definitions are often so varied and vague that not even their various proponents agree on them, like logical behaviorism, philosophical behaviorism, methodological behaviorism, radical behaviorism, neobehaviorism, and many more. By “intentionality,” it discusses the Brentano-Husserl notion of the directedness of the mental. The directedness of the mental simply meant that for each mental phenomenon. In studying man, one needs a very comprehensive theory that includes a theory of action, a theory of meaning and communication, a theory of reference, and a theory of knowledge. All of these are interconnected in such a way that evidence is transmitted between the theories.
Convention was immediately recognized as a major contribution to the subject and its significance has remained undiminished since its first publication in 1969. Lewis analyzes social conventions as regularities in the resolution of recurring coordination problems-situations characterized by interdependent decision processes in which common interests are at stake. Conventions are contrasted with other kinds of regularity, and conventions governing systems of communication are given special attention.