Greta Warren

Greta Warren, Mark T Keane, Ruth M J Byrne

Counterfactual explanations are increasingly used to address interpretability, recourse, and bias in AI decisions. However, we do not know how well counterfactual explanations help users to understand a systems decisions, since no large scale user studies have compared their efficacy to other sorts of explanations such as causal explanations (which have a longer track record of use in rule based and decision tree models). It is also unknown whether counterfactual explanations are equally effective for categorical as for continuous features, although current methods assume they do. Hence, in a controlled user study with 127 volunteer participants, we tested the effects of counterfactual and causal explanations on the objective accuracy of users predictions of the decisions made by a simple AI system, and participants subjective judgments of satisfaction and trust in the explanations. We discovered a dissociation between objective and subjective measures: counterfactual explanations elicit higher accuracy of predictions than no-explanation control descriptions but no higher accuracy than causal explanations, yet counterfactual explanations elicit greater satisfaction and trust than causal explanations. We also found that users understand explanations referring to categorical features more readily than those referring to continuous features. We discuss the implications of these findings for current and future counterfactual methods in XAI.

Greta Warren, Mark T. Keane, Christophe Gueret et al.

Counterfactual explanations are an increasingly popular form of post hoc explanation due to their (i) applicability across problem domains, (ii) proposed legal compliance (e.g., with GDPR), and (iii) reliance on the contrastive nature of human explanation. Although counterfactual explanations are normally used to explain individual predictive-instances, we explore a novel use case in which groups of similar instances are explained in a collective fashion using ``group counterfactuals'' (e.g., to highlight a repeating pattern of illness in a group of patients). These group counterfactuals meet a human preference for coherent, broad explanations covering multiple events/instances. A novel, group-counterfactual algorithm is proposed to generate high-coverage explanations that are faithful to the to-be-explained model. This explanation strategy is also evaluated in a large, controlled user study (N=207), using objective (i.e., accuracy) and subjective (i.e., confidence, explanation satisfaction, and trust) psychological measures. The results show that group counterfactuals elicit modest but definite improvements in people's understanding of an AI system. The implications of these findings for counterfactual methods and for XAI are discussed.

Will Hawkins, Kaivalya Rawal, Jonathan Rystrøm et al.

Fine-tuning a large language model is a ubiquitous method for enhancing its capability on a specific downstream task. However, prior work has shown that this increase in capability comes with a cost: it can increase a model's tendency to respond to unsafe adversarial prompts, even when fine-tuning with non-adversarial data. We present the first comprehensive empirical study of this phenomenon in multilingual settings by fine-tuning Llama-3.2, Qwen3, and Gemma-3 models using benign data translated across nine languages. We find that safety outcomes are highly sensitive to both the choice of fine-tuning language and the evaluation language, with adversarial compliance rates increasing four-fold in some settings. Multilingual safety drift is decoupled from general capability metrics, and occurs heterogeneously across languages and models. Fine-tuning in non-English languages often induces smaller internal representational drifts than English, but these shifts lead models to default to either exaggerated compliance or refusal. As such, assessing fine-tuning impacts solely in English provides inadequate assurance for deployment. To facilitate further research into these cross-lingual safety blind spots, we release the Multilingual-Benign-Tune dataset and the SORRY-Bench-Multilingual evaluation suite.

Greta Warren, Irina Shklovski, Isabelle Augenstein

The pervasiveness of large language models and generative AI in online media has amplified the need for effective automated fact-checking to assist fact-checkers in tackling the increasing volume and sophistication of misinformation. The complex nature of fact-checking demands that automated fact-checking systems provide explanations that enable fact-checkers to scrutinise their outputs. However, it is unclear how these explanations should align with the decision-making and reasoning processes of fact-checkers to be effectively integrated into their workflows. Through semi-structured interviews with fact-checking professionals, we bridge this gap by: (i) providing an account of how fact-checkers assess evidence, make decisions, and explain their processes; (ii) examining how fact-checkers use automated tools in practice; and (iii) identifying fact-checker explanation requirements for automated fact-checking tools. The findings show unmet explanation needs and identify important criteria for replicable fact-checking explanations that trace the model's reasoning path, reference specific evidence, and highlight uncertainty and information gaps.