Search

In this paper we describe an argumentation-based representation of normal form games, and demonstrate how argumentation can be used to compute pure strategy Nash equilibria. Our approach builds on Modgil’s Extended Argumentation Frameworks. We demonstrate its correctness, showprove several theoretical properties it satisfies, and outline how it can be used to explain why certain strategies are Nash equilibria to a non-expert human user.

In this paper we illustrate how novel AI methods can improve the performance of intelligence analysts. These analysts aim to make sense of — often conflicting or incomplete — information, weighing up competing hypotheses which serve to explain an observed situation. Analysts have access to numerous visual analytic tools which support the temporal and/or conceptual structuring of information and collection, and support the evaluation of alternative hypotheses. We believe, however, that there are currently no tools or methods which allow analysts to combine the recording and interpretation of information, and that there is little understanding about how software tools can facilitate the hypothesis formation process. Following the identification of these requirements, we developed the CISpaces (Collaborative Intelligence Spaces) decision support tool in collaboration with professional intelligence analysts. CISpaces combines multiple AI-based methods including argumentation theory, crowdsourced Bayesian analysis, and provenance recording. We show that CISpaces is able to provide support to analysts by facilitating the interpretation of different types of evidence through argumentation-based reasoning, provenance analysis and crowdsourcing. We undertook an experimental analysis with intelligence analysts which highlights three key points. (1) The novel, principled AI methods implemented in CISpaces advance performance in intelligence analysis. (2) While designed as a research prototype, analysts benchmarked it against their existing software tools, and we provide results suggesting intention to adopt CISpaces in analysts’ daily activities. (3) Finally, the evaluation highlights some drawbacks in CISpaces. However, these are not due to the technologies underpinning the tool, but rather in its lack of integration with existing organisational standards regarding input and output formats. Our evaluation with intelligence analysts therefore demonstrates the potential impact that an integrated tool building on state-of-the-art AI techniques can have on the process of understanding complex situations, and on how such a tool can help focus human effort on identifying more credible interpretations of evidence.

Abstract If artificial intelligence (AI) is to help solve individual, societal and global problems, humans should neither underestimate nor overestimate its trustworthiness. Situated in-between these two extremes is an ideal ‘Goldilocks’ zone of credibility. But what will keep trust in this zone? We hypothesise that this role ultimately falls to the social cognition mechanisms which adaptively regulate conformity between humans. This novel hypothesis predicts that human-like functional biases in conformity should occur during interactions with AI. We examined multiple tests of this prediction using a collaborative remembering paradigm, where participants viewed household scenes for 30 s vs. 2 min, then saw 2-alternative forced-choice decisions about scene content originating either from AI- or human-sources. We manipulated the credibility of different sources (Experiment 1) and, from a single source, the estimated-likelihood (Experiment 2) and objective accuracy (Experiment 3) of specific decisions. As predicted, each manipulation produced functional biases for AI-sources mirroring those found for human-sources. Participants conformed more to higher credibility sources, and higher-likelihood or more objectively accurate decisions, becoming increasingly sensitive to source accuracy when their own capability was reduced. These findings support the hypothesised role of social cognition in regulating AI’s influence, raising important implications and new directions for research on human–AI interaction.