No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
An Axiomatic Approach to Formalized Responsibility Ascription
Abstract
A formalized and quantifiable responsibility score is a crucial component in many aspects of the development and application of multi-agent systems and autonomous agents. We can employ it to inform decision making processes based on ethical considerations, as a measure to ensure redundancy that helps us in avoiding system failure, as well as for verifying that autonomous systems remain trustworthy by testing for unwanted responsibility voids in advance. We follow recent proposals to use probabilities as the basis for responsibility ascription in uncertain environments rather than the deterministic causal views employed in much of the previous formal philosophical literature. Using an axiomatic approach we formally evaluate the qualities of (classes of) proposed responsibility functions. To this end, we decompose the computation of the responsibility a group carries for an outcome into the computation of values that we assign to its members for individual decisions leading to that outcome, paired with an appropriate aggregation function. Next, we discuss a number of intuitively desirable properties for each of these contributing functions. We find an incompatibility between axioms determining upper and lower bounds for the values assigned at the member level. Regarding the aggregation from member-level values to group-level responsibility we are able to axiomatically characterise one promising aggregation function. Finally, we present two maximally axiom compliant group-level responsibility measures – one respecting the lower bound axioms at the member level and one respecting the corresponding upper bound axioms.
ResearchGate has not been able to resolve any citations for this publication.
p>A central issue in the specification and verification of autonomous agents and multiagent systems is the ascription of responsibility to individual agents and groups of agents When designing a (multi)agent system, we must specify which agents or groups of agents are responsible for bringing about a particular state of affairs Similarly, when verifying a multiagent system, we may wish to determine the responsibility of agents or groups of agents for a particular state of affairs, and the contribution of each agent to bringing about that state of affairs In this paper, we discuss several aspects of responsibility, including strategic ability of agents, their epistemic properties, and their relationship to the evolution of the system behavior We introduce a formal framework for reasoning about the responsibility of individual agents and agent groups in terms of the agents' strategies and epistemic properties, and state some properties of the framework.</p
The present article is devoted to a logical treatment of some fundamental concepts involved in responsibility attribution. We specify a theoretical framework based on a language of temporal deontic logic with agent-relative operators for deliberate causal contribution. The framework is endowed with a procedure to solve normative conflicts which arise from the assessment of different normative sources. We provide a characterization result for a basic system within this framework and illustrate how the concepts formalized can be put at work in the analysis of examples of legal reasoning.
In this paper, we develop a formalisation of the main ideas of the work of Van de Poel on responsibility. Using the basic concepts through which the meanings of responsibility are defined, we construct a logic which enables to express sentences like “individual i is accountable for φ”, “individual i is blameworthy for φ” and “individual i has the obligation to see to it that φ”. This formalization clarifies the definitions of responsibility given by Van de Poel and highlights their differences and similarities. It also helps to assess the consistency of the formalisation of responsibility, not only by showing that definitions are not inconsistent, but also by providing a formal demonstration of the relation between three main meanings of responsibility (accountability, blameworthiness, and obligation). The formal account can be used to derive new properties of the concepts. With the help of the formalisation, we detect the occurrence of the problem of many hands (PMH) by defining a logical framework for reasoning about collective and individual responsibility. This logic extends the Coalition Epistemic Dynamic Logic (CEDL) by adding a notion of group knowledge (and generalize the definitions of individual responsibility to groups of agents), agent ability and knowing how to its semantics.
Individualists claim that collective obligations are reducible to the individual obligations of the collective’s members. Collectivists deny this. We set out to discover who is right by way of a deontic logic of collective action that models collective actions, abilities, obligations, and their interrelations. On the basis of our formal analysis, we argue that when assessing the obligations of an individual agent, we need to distinguish individual obligations from member obligations. If a collective has a collective obligation to bring about a particular state of affairs, then it might be that no individual in the collective has an individual obligation to bring about that state of affairs. What follows from a collective obligation is that each member of the collective has a member obligation to help ensure that the collective fulfills its collective obligation. In conclusion, we argue that our formal analysis supports collectivism.
Do responsibility voids exist? That is, are there situations in which the group is collectively morally responsible for some outcome although no member can be held individually morally responsible for it? To answer these questions, I draw a distinction between competitive and cooperative decision contexts based on the team-reasoning account of cooperation. Accordingly, I provide a reasoning-based analysis of cooperation, competition, moral responsibility, and, last, potential responsibility voids. I then argue that competitive decision contexts are free of responsibility voids. The conditions for the existence of responsibility voids in cooperative decision contexts depend on the type of uncertainty the group faces, either external or coordination uncertainty.
Debates on lethal autonomous weapon systems have proliferated in the past 5 years. Ethical concerns have been voiced about a possible raise in the number of wrongs and crimes in military operations and about the creation of a “responsibility gap” for harms caused by these systems. To address these concerns, the principle of “meaningful human control” has been introduced in the legal–political debate; according to this principle, humans not computers and their algorithms should ultimately remain in control of, and thus morally responsible for, relevant decisions about (lethal) military operations. However, policy-makers and technical designers lack a detailed theory of what “meaningful human control” exactly means. In this paper, we lay the foundation of a philosophical account of meaningful human control, based on the concept of “guidance control” as elaborated in the philosophical debate on free will and moral responsibility. Following the ideals of “Responsible Innovation” and “Value-sensitive Design,” our account of meaningful human control is cast in the form of design requirements. We identify two general necessary conditions to be satisfied for an autonomous system to remain under meaningful human control: first, a “tracking” condition, according to which the system should be able to respond to both the relevant moral reasons of the humans designing and deploying the system and the relevant facts in the environment in which the system operates; second, a “tracing” condition, according to which the system should be designed in such a way as to grant the possibility to always trace back the outcome of its operations to at least one human along the chain of design and operation. As we think that meaningful human control can be one of the central notions in ethics of robotics and AI, in the last part of the paper, we start exploring the implications of our account for the design and use of non-military autonomous systems, for instance, self-driving cars.
The aim of this article is to provide a logical analysis of the concept of responsibility attribution; that is, how agents
ascribe responsibility about the consequences of actions, either to themselves or to other agents. The article is divided
in two parts. The first part investigates the importance of the concept of responsibility attribution for emotion theory in
general and, in particular, for the theory of attribution emotions such as guilt, pride, moral approval and moral disapproval.
The second part explores the collective dimension of responsibility attribution and attribution emotions, namely the concepts
of collective responsibility and collective guilt. The proposed analysis is based on an extension of the logic STIT (the logic
of ‘Seeing To It That’) with three different types of knowledge and common knowledge modal operators depending on the time
of choice: before one’s choice, after one’s choice but before knowing the choices of other agents, and after the choices of
all agents have become public. Decidability of the satisfiability problem of the logic is studied in the article.
Several local and global properties of (extended) aggregation functions are discussed and their relationships are examined.
Some special classes of averaging, conjunctive and disjunctive aggregation functions are reviewed. A special attention is
paid to the weighted aggregation functions, including some construction methods
This paper distinguishes six different responsibility concepts from one another, and it explains how those concepts relate to each other. The resulting “structured taxonomy of responsibility concepts” identifies several common sources of disputes about responsibility, and it suggests a procedure for resolving such disputes. To demonstrate their utility, this taxonomy and procedure are then used to illuminate debates in two familiar contexts. AVAILABLE ONLINE.
The aim of this work is to propose a logical framework for the specification of cognitive emotions that are based on counterfactual reasoning about agents' choices. The prototypical counterfactual emotion is regret. In order to meet this objective, we exploit the well-known STIT logic (Belnap et al. (2001) [9], Horty (2001) [30], Horty and Belnap (1995) [31]). STIT logic has been proposed in the domain of formal philosophy in the nineties and, more recently, it has been imported into the field of theoretical computer science where its formal relationships with other logics for multi-agent systems such as ATL and Coalition Logic (CL) have been studied. STIT is a very suitable formalism to reason about choices and capabilities of agents and groups of agents. Unfortunately, the version of STIT with agents and groups has been recently proved to be undecidable and not finitely axiomatizable. In this work we study a decidable and finitely axiomatizable fragment of STIT with agents and groups which is sufficiently expressive for our purpose of formalizing counterfactual emotions. We call dfSTITdfSTIT our STIT fragment. After having extended dfSTITdfSTIT with knowledge modalities, in the second part of article, we exploit it in order to formalize four types of counterfactual emotions: regret, rejoicing, disappointment, and elation. At the end of the article we present an application of our formalization of counterfactual emotions to a concrete example.
One way to allocate tasks to agents is by ascribing them obligations. From obligations to be, agents are able to infer what
are the forbidden, permitted and obligatory actions they may perform, by using the well-known Meyer’s reduction from obligations
to be to obligations to do. However, we show through an example that this method is not completely adequate to guide agents’
decisions. We then propose a solution using, instead of obligations, the concept of ‘responsibility’. To formalise responsibility
we use a multiagent extension of propositional dynamic logic as framework, and then we define some basic concepts, such as
‘agent ability’, also briefly discussing the problem of uniform strategies and a possible solution. In the last part, we show
that our framework can be used in the specification of normative multiagent systems, by presenting an extensive running example.
This is a study of the axiomatic method and its recent applications to game theory and resource allocation. It begins with a user's guide. This guide first describes the components of an axiomatic study, discusses the logical and conceptual independence of the axioms in a characterization, exposes mistakes that are often made in the formulation of axioms, and emphasizes the importance of seeing each axiomatic study from the perspective of the axiomatic program. It closes with a schematic presentation of this program. The second part of this study discusses the scope of the axiomatic method and briefly presents a number of models where its use have been particularly successful. It presents alternatives to the axiomatic method and answers criticisms often addressed at the axiomatic method. It delimits the scope of the method and illustrates its relevance to the study of resource allocation and the study of strategic interaction. Finally, it provides extensive illustrations of the considerable recent success that the method has met in the study of a number of new models.
In this elucidating work, the authors attempt to construct a logical analysis of human actions, focusing on those actions based on choice. Using their examination of "seeing to it that," they investigate a large number of topics, including imperatives, deontic logic, strategies, determinism, and promising assertion. One of the work's provocative conclusions is that one, actual future does not exist; instead, all possible futures are on par with one another.
We provide formal definitions of degree of blameworthiness and intention relative to an epistemic state (a probability over causal models and a utility function on outcomes). These, together with a definition of actual causality, provide the key ingredients for moral responsibility judgments. We show that these definitions give insight into commonsense intuitions in a variety of puzzling cases from the literature.
We propose a refinement of STIT logic to make it suitable to model causal agency and responsibility in basic multi-agent scenarios in which agents can interfere with one another. We do this by supplementing STIT semantics, first, with action types and, second, with a relation of opposing between action types. We exploit these novel elements to represent a test for potential causation, based on an intuitive notion of expected result of an action, and two tests for actual causation from the legal literature, i.e., the but-for and the NESS tests. We then introduce three new STIT operators modeling corresponding notions of causal responsibility, which we call potential, strong, and plain responsibility, and use them to provide a fine-grained analysis of a number of case studies involving both individual agents and groups.
It is often natural to compare two events by describing one as ‘more of a cause’ of some effect than the other. But what do such comparisons amount to, exactly? This paper aims to provide a guided tour of the recent literature on ‘degrees of causation’. Section 2 looks at what I call ‘dependence measures’, which arise from thinking of causes as difference-makers. Section 3 looks at what I call ‘production measures’, which arise from thinking of causes as jointly sufficient for their effects. Finally, section 4 examines the important question of whether there is any sense in which an agent is more responsible for an outcome in virtue of her action being more of a cause of it. I describe a puzzle that emerges from this question, and explore various strategies for resolving it.
Institutional rules create difficulties for the allocation of moral responsibility. One problem is the existence of responsibility voids, i.e. situations in which an outcome results from individual interactions but for which no one is responsible. Another is that responsibility can be fragmented in the sense that responsibility-bearing individuals may be responsible for different features of the outcome. This study examines both problems together. We show that for a large class of situations the two problems are logically dependent. More precisely, non-dictatorial decision procedures can only ensure the absence of voids if they allow for the fragmentation of responsibility.
“Probabilistic Causation” designates a group of theories that aim to characterize the relationship between cause and effect using the tools of probability theory. The central idea behind these theories is that causes change the probabilities of their effects. This article traces developments in probabilistic causation, including recent developments in causal modeling. A variety of issues within, and objections to, probabilistic theories of causation will also be discussed.
Face au phenomene de causation redondante qui l'oblige a reviser son analyse contrefactuelle de la causation, l'A. met en evidence la distinction entre la cause (preempting) et l'effet (preempted) de la causation preemptive, quelque soit le cas, symetrique ou asymetrique, de la preemption premiere et de la preemption tardive, definies comme rupture (cutting) de la chaine causale. Rejetant l'idee de quasi-dependance, et corrigeant celle de fragilite, l'A. definit la transitivite de la causation comme modele d'influence.
This contribution discusses the relation between forward-looking and backward-looking responsibility. The notion of forward-looking responsibility I focus on is, following Robert Goodin, that of seeing to it that a certain state of affairs obtains. In addition, I focus on two types of backward-looking responsibility: accountability and blameworthiness. I argue that accountability only entails blameworthiness if the agent cannot cite certain reasonable excuses like ignorance and compulsion (which were already mentioned by Aristotle). It is further argued that accountability can both be based on not properly discharging a forward-looking responsibility and on the breach of a duty that caused a certain negative consequence. I show that in both cases three general conditions need to apply in order to hold an agent reasonably accountable: a capacity condition, a causality condition and a wrongdoing condition. The exact content of these conditions, however, depends on whether accountability is based on a forward-looking responsibility that is not properly discharged or on the breach of a duty.
The concept of agent-responsibility for an outcome (that is, of the outcome reflecting the autonomous choice of the agent) is central to both ethics and political philosophy. The concept, however, remains radically under-explored. In particular, the issue of partial responsibility for an outcome needs further development. I propose an account of partial responsibility based on partial causal contribution. Agents who choose autonomously in full knowledge of the consequences are agent-responsible, I claim, for the shift in the objective probability of the outcome in question that her choice induces. Thus, agents will typically be only partially agent-responsible (that is, for a shift of less than 100 percent) for any given outcome. The model has an implication that is generally rejected: that agents who purchase lottery tickets and win are agent-responsible for only part of the winnings.
Develops deontic logic against the background of a rigorous theory of agency in branching, or indeterministic, time. It is often assumed that the notion of what an agent ought to do can be identified with that of what it ought to be that the agent does. The book provides a framework in which this assumption can be formulated precisely and shown to be mistaken. In its place, it offers an alternative account of what agents ought to do that relies on an analogy between action in indeterministic time and choice under uncertainty, as it is studied in decision theory. This alternative account is then related to issues involving conditional obligation, group obligation, act utilitarianism, and rule utilitarianism.
In this paper I start by briefly presenting an analysis of token cause and of token causal relevance that I developed elsewhere, and then apply it to the famous thirsty traveler riddle. One general outcome of the analysis of causal relevance employed here is that in preemption cases (early or late) the preempted cause is not a cause since it is causally irrelevant to the effect. I consider several variations of the thirsty traveler riddle. In the first variation the first enemy emptied the canteen and the second enemy threw it away. On this variation, the act of neither enemy comes out, on the analysis employed here, as causally relevant to, and thus not as a cause of, the fact that the traveler died, but the conjunction of the two acts is a cause of it. This version is a case of mutual preemption. I argue that it has the same structure as the voting paradox, which thus has an analogous solution. In the standard version, in which the first enemy added poison to the water in the traveler's canteen, the act of the second enemy (who threw the canteen away) comes out, on the analysis used here, as causally relevant to and as a cause of the fact that the traveler died, but the act of the first enemy comes out as neither. I also make a comparison with Lewis' accounts, and discuss alternative treatments of the puzzle such as those of Hart and Honore and of Gavison, Margalit, and Ullmann-Margalit.
We define an extension of stit logic that encompasses subjective probabilities representing beliefs about simultaneous choice exertion of other agents. The formalism enables us to express the notion of 'attempt' as a choice exertion that maximizes the chance of success with respect to an action effect. The notion of attempt (or effort) is central in philosophical and legal discussions on responsibility and liability.
We propose a new definition of actual causes, using structural equations to model counterfactuals. We show that the definition yields a plausible and elegant account of causation that handles well
examples which have caused problems for other definitions and resolves major difficulties in the traditional account.
Introduction
Causal models: a review
2.1 Causal models
2.2 Syntax and semantics
The definition of cause
Examples
A more refined definition
Discussion
AAppendix: Some Technical Issues
A.1 The active causal process
A.2 A closer look at AC2(b)
A.3 Causality with infinitely many variables
A.4 Causality in nonrecursive models
Influence and responsibility: a logical analysis
- E Lorini
- G Sartor
Causation in the law
- M Moore
Causation, responsibility, risk, probability, naked statistics, and proof: pruning the bramble bush by clarifying the concepts
- R W Wright
- RW Wright
Accountability and responsibility in agent organizations
- M Baldoni
- C Baroglio
- O Boissier
- K M May
- R Micalizio
- S Tedeschi
An anatomy of moral responsibility
- M Braham
- M Van Hees