Accountability for Practical Reasoning Agents

Cranefield, Stephen; Oren, Nir; Vasconcelos, Wamberto

doi:10.1007/978-3-030-17294-7_3

Cited by 12 publications

(8 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Last but not least, explainability (EX), i.e., the capability of a model to provide an explanation of its processes [9], is desirable in any autonomous system for reasons of trustworthiness [18], accountability [19], and responsibility [20]; this is particularly crucial in the MPC context for allowing users to know why a solution is suggested and its effects [8], and to align the differences between uploaders and co-owners [5].…”

Section: Requirements For Mpc Solutionsmentioning

confidence: 99%

An explainable assistant for multiuser privacy

Mosca

Such

2022

Auton Agent Multi-Agent Syst

View full text Add to dashboard Cite

Multiuser Privacy (MP) concerns the protection of personal information in situations where such information is co-owned by multiple users. MP is particularly problematic in collaborative platforms such as online social networks (OSN). In fact, too often OSN users experience privacy violations due to conflicts generated by other users sharing content that involves them without their permission. Previous studies show that in most cases MP conflicts could be avoided, and are mainly due to the difficulty for the uploader to select appropriate sharing policies. For this reason, we present ELVIRA, the first fully explainable personal assistant that collaborates with other ELVIRA agents to identify the optimal sharing policy for a collectively owned content. An extensive evaluation of this agent through software simulations and two user studies suggests that ELVIRA, thanks to its properties of being role-agnostic, adaptive, explainable and both utility- and value-driven, would be more successful at supporting MP than other approaches presented in the literature in terms of (i) trade-off between generated utility and promotion of moral values, and (ii) users’ satisfaction of the explained recommended output.

show abstract

Section: Requirements For Mpc Solutionsmentioning

confidence: 99%

An explainable assistant for multiuser privacy

Mosca

Such

2022

Auton Agent Multi-Agent Syst

View full text Add to dashboard Cite

show abstract

“…In [17], Cranefield et al discuss the notion of accountable autonomy within the context of practical reasoning agents. The authors start by listing some requirements that an agent should satisfy to be capable of performing practical reasoning on accountability.…”

Section: Related Workmentioning

confidence: 99%

MOCA: An ORM model for computational accountability

Baldoni

Baroglio

May

et al. 2019

View full text Add to dashboard Cite

Accountability is often seen as a key notion in distributed systems, both in the human world and in software, on top of which interaction is built, together with the sibling notion of responsibility. However, there is limited support for the specification and use of such concepts in computational settings. This paper proposes an information model for computational accountability, which describes what data have to be available to allow, in any situation of interest arising from a group of interacting parties, the investigation and identification of accountability. The characterization of accountability provided by the model is grounded in the notions of just expectation and control. The information model is expressed in Object-Role Modeling (ORM), since it is well-suited to capture the relational nature of the accountability concept. An advantage of our model is that a designer can verify the consistency of a domain w.r.t. a set of accountability requirements. The paper exemplifies this checking by means of an Answer Set Program (ASP).

show abstract

“…An experience questionnaire (extracted and modified from GEQ [14]) was given to each participant at the end of each round. We clustered questions in three main groups (GEQ indices in brackets): Competence (10,15,17,21); Affect (9,22,24); and Challenge (23,26,33). We included four custom questions to evaluate game-specific criteria, such as how often they consulted the text rules and if they anticipated/agreed with agents' actions.…”

Section: User Studymentioning

confidence: 99%

“…Their approach is formally rigorous but is specialised for a specific set of traffic rules only and does not generalise beyond. Cranefield et al [9] propose that ideal accountable agents must: i) understand what is expected from them (from rules/obligations); ii) answer queries about their decision-making (being explainable); iii) carry out argumentative dialogues in which beliefs and plans are challenged and justified; iv) adapt their reasoning apparatuses or update their plans as a result of accountability dialogues; and v) take human values into account when reasoning.…”

Section: Introductionmentioning

confidence: 99%

“…Argumentation approaches walk hand in hand with the desiderata proposed by Cranefield et al [9], as they allow us to: i) enable norm-aware reasoning [4]; ii) generate explanations [11,25]; iii) carry argumentative dialogues to support their positions [1,23]; iv) perform meta-reasoning [26]; and v) consider human values [16]. Consequently, we regard argumentation frameworks as a strong mechanism for providing accountable and explainable agency.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Culture-Based Explainable Human-Agent Deconfliction

Raymond,

Gunes,

Prorok

2019

Preprint

View full text Add to dashboard Cite

Law codes and regulations help organise societies for centuries, and as AI systems gain more autonomy, we question how human-agent systems can operate as peers under the same norms, especially when resources are contended. We posit that agents must be accountable and explainable by referring to which rules justify their decisions. The need for explanations is associated with user acceptance and trust. This paper's contribution is twofold: i) we propose an argumentation-based human-agent architecture to map human regulations into a culture for artificial agents with explainable behaviour. Our architecture leans on the notion of argumentative dialogues and generates explanations from the history of such dialogues; and ii) we validate our architecture with a user study in the context of human-agent path deconfliction. Our results show that explanations provide a significantly higher improvement in human performance when systems are more complex. Consequently, we argue that the criteria defining the need of explanations should also consider the complexity of a system. Qualitative findings show that when rules are more complex, explanations significantly reduce the perception of challenge for humans.

show abstract

Accountability for Practical Reasoning Agents

Cited by 12 publications

References 37 publications

An explainable assistant for multiuser privacy

An explainable assistant for multiuser privacy

MOCA: An ORM model for computational accountability

Culture-Based Explainable Human-Agent Deconfliction

Contact Info

Product

Resources

About