Ofra Amir

Hana Matatov, Mor Naaman, Ofra Amir

The massive spread of visual content through the web and social media poses both challenges and opportunities. Tracking visually-similar content is an important task for studying and analyzing social phenomena related to the spread of such content. In this paper, we address this need by building a dataset of social media images and evaluating visual near-duplicates retrieval methods based on image retrieval and several advanced visual feature extraction methods. We evaluate the methods using a large-scale dataset of images we crawl from social media and their manipulated versions we generated, presenting promising results in terms of recall. We demonstrate the potential of this method in two case studies: one that shows the value of creating systems supporting manual content review, and another that demonstrates the usefulness of automatic large-scale data analysis.

Yael Septon, Tobias Huber, Elisabeth André et al.

Explaining the behavior of reinforcement learning agents operating in sequential decision-making settings is challenging, as their behavior is affected by a dynamic environment and delayed rewards. Methods that help users understand the behavior of such agents can roughly be divided into local explanations that analyze specific decisions of the agents and global explanations that convey the general strategy of the agents. In this work, we study a novel combination of local and global explanations for reinforcement learning agents. Specifically, we combine reward decomposition, a local explanation method that exposes which components of the reward function influenced a specific decision, and HIGHLIGHTS, a global explanation method that shows a summary of the agent's behavior in decisive states. We conducted two user studies to evaluate the integration of these explanation methods and their respective benefits. Our results show significant benefits for both methods. In general, we found that the local reward decomposition was more useful for identifying the agents' priorities. However, when there was only a minor difference between the agents' preferences, then the global information provided by HIGHLIGHTS additionally improved participants' understanding.

Yotam Amitai, Guy Avni, Ofra Amir

As reinforcement learning methods increasingly amass accomplishments, the need for comprehending their solutions becomes more crucial. Most explainable reinforcement learning (XRL) methods generate a static explanation depicting their developers' intuition of what should be explained and how. In contrast, literature from the social sciences proposes that meaningful explanations are structured as a dialog between the explainer and the explainee, suggesting a more active role for the user and her communication with the agent. In this paper, we present ASQ-IT -- an interactive tool that presents video clips of the agent acting in its environment based on queries given by the user that describe temporal properties of behaviors of interest. Our approach is based on formal methods: queries in ASQ-IT's user interface map to a fragment of Linear Temporal Logic over finite traces (LTLf), which we developed, and our algorithm for query processing is based on automata theory. User studies show that end-users can understand and formulate queries in ASQ-IT, and that using ASQ-IT assists users in identifying faulty agent behaviors.

Dana Harari, Ofra Amir

Artificial intelligence (AI) assistants are increasingly embedded in workplace tools, raising the question of how initiative-taking shapes adoption. Prior work highlights trust and expectation mismatches as barriers, but the underlying psychological mechanisms remain unclear. Drawing on self-affirmation and social exchange theories, we theorize that unsolicited help elicits self-threat, thereby reducing willingness to accept help, likelihood of future use, and performance expectancy of AI. We report two vignette-based experiments (Study~1: $N=761$; Study~2: $N=571$, preregistered). Study~1 compared anticipatory and reactive help provided by an AI vs. a human, while Study~2 distinguished between \emph{offering} (suggesting help) and \emph{providing} (acting automatically). In Study 1, AI reactive help was more threatening than reactive human help. Across both studies, anticipatory help increased user's self-threat and reduced adoption outcomes. Our findings identify self-threat as a mechanism through which anticipatory help, a proactive AI feature, may backfire, and suggest design implications to be tested in interactive systems.

Bar Eini Porat, Rom Gutman, Uri Shalit et al.

Explainability methods have progressed rapidly, but global explanations for time-series models remain underdeveloped, with most approaches focusing on local, instance-level attributions. We introduce INSIGHTS, a model-agnostic, user-centric approach for providing global explanations of time series models. Our approach prioritizes simplicity, efficiency, and transparency in its design, ensuring that stakeholders can readily adopt its outputs. While current methods focus on local explanations, INSIGHTS generates sample summaries that offer a comprehensive overview of model behavior. It balances the importance and diversity of time series samples to create informative subsets using utility functions that capture domain-specific aspects of time series behavior, such as exceeding domain norms. We evaluate INSIGHTS through experiments, interviews, and a user study. Our results indicate INSIGHTS effectively constructs comprehensive, diverse time series subsets, producing summaries manageable for individual evaluation. It is preferred by domain experts for its ability to provide a stable understanding of model behavior and the quality of the samples identified. Moreover, user study participants presented with INSIGHTS-based summaries exhibit an enhanced understanding of the model's overall behavior.

Zachary Bamberger, Till R. Saenger, Gilad Morad et al.

Inference-Time-Compute (ITC) methods like Best-of-N and Tree-of-Thoughts are meant to produce output candidates that are both high-quality and diverse, but their use of high-temperature sampling often fails to achieve meaningful output diversity. Moreover, existing ITC methods offer limited control over how to perform reasoning, which in turn limits their explainability. We present STATe-of-Thoughts (STATe), an interpretable ITC method that searches over high-level reasoning patterns. STATe replaces stochastic sampling with discrete and interpretable textual interventions: a controller selects actions encoding high-level reasoning choices, a generator produces reasoning steps conditioned on those choices, and an evaluator scores candidates to guide search. This structured approach yields three main advantages. First, action-guided textual interventions produce greater response diversity than temperature-based sampling. Second, in a case study on argument generation, STATe's explicit action sequences capture interpretable features that are highly predictive of output quality. Third, estimating the association between performance and action choices allows us to identify promising yet unexplored regions of the action space and steer generation directly toward them. Together, these results establish STATe as a practical framework for generating high-quality, diverse, and interpretable text. Our framework is available at https://github.com/zbambergerNLP/state-of-thoughts.

Yotam Amitai, Yael Septon, Ofra Amir

Explainable reinforcement learning (XRL) methods aim to help elucidate agent policies and decision-making processes. The majority of XRL approaches focus on local explanations, seeking to shed light on the reasons an agent acts the way it does at a specific world state. While such explanations are both useful and necessary, they typically do not portray the outcomes of the agent's selected choice of action. In this work, we propose ``COViz'', a new local explanation method that visually compares the outcome of an agent's chosen action to a counterfactual one. In contrast to most local explanations that provide state-limited observations of the agent's motivation, our method depicts alternative trajectories the agent could have taken from the given state and their outcomes. We evaluated the usefulness of COViz in supporting people's understanding of agents' preferences and compare it with reward decomposition, a local explanation method that describes an agent's expected utility for different actions by decomposing it into meaningful reward types. Furthermore, we examine the complementary benefits of integrating both methods. Our results show that such integration significantly improved participants' performance.

Yotam Amitai, Ofra Amir, Guy Avni

As reinforcement learning methods increasingly amass accomplishments, the need for comprehending their solutions becomes more crucial. Most explainable reinforcement learning (XRL) methods generate a static explanation depicting their developers' intuition of what should be explained and how. In contrast, literature from the social sciences proposes that meaningful explanations are structured as a dialog between the explainer and the explainee, suggesting a more active role for the user and her communication with the agent. In this paper, we present ASQ-IT -- an interactive explanation system that presents video clips of the agent acting in its environment based on queries given by the user that describe temporal properties of behaviors of interest. Our approach is based on formal methods: queries in ASQ-IT's user interface map to a fragment of Linear Temporal Logic over finite traces (LTLf), which we developed, and our algorithm for query processing is based on automata theory. User studies show that end-users can understand and formulate queries in ASQ-IT and that using ASQ-IT assists users in identifying faulty agent behaviors.

Sahar Admoni, Ofra Amir, Assaf Hallak et al.

Large language models (LLMs) seem to offer an easy path to interpretability: just ask them to explain their decisions. Yet, studies show that these post-hoc explanations often misrepresent the true decision process, as revealed by mismatches in feature importance. Despite growing evidence of this inconsistency, no systematic solutions have emerged, partly due to the high cost of estimating feature importance, which limits evaluations to small datasets. To address this, we introduce the Post-hoc Self-Consistency Bank (PSCB) - a large-scale benchmark of decisions spanning diverse tasks and models, each paired with LLM-generated explanations and corresponding feature importance scores. Analysis of PSCB reveals that self-consistency scores barely differ between correct and incorrect predictions. We also show that the standard metric fails to meaningfully distinguish between explanations. To overcome this limitation, we propose an alternative metric that more effectively captures variation in explanation quality. We use it to fine-tune LLMs via Direct Preference Optimization (DPO), leading to significantly better alignment between explanations and decision-relevant features, even under domain shift. Our findings point to a scalable path toward more trustworthy, self-consistent LLMs.

Yotam Amitai, Reuth Mirsky, Ofra Amir

In human-agent teams, openly sharing goals is often assumed to enhance planning, collaboration, and effectiveness. However, direct communication of these goals is not always feasible, requiring teammates to infer their partner's intentions through actions. Building on this, we investigate whether an AI agent's ability to share its inferred understanding of a human teammate's goals can improve task performance and perceived collaboration. Through an experiment comparing three conditions-no recognition (NR), viable goals (VG), and viable goals on-demand (VGod) - we find that while goal-sharing information did not yield significant improvements in task performance or overall satisfaction scores, thematic analysis suggests that it supported strategic adaptations and subjective perceptions of collaboration. Cognitive load assessments revealed no additional burden across conditions, highlighting the challenge of balancing informativeness and simplicity in human-agent interactions. These findings highlight the nuanced trade-off of goal-sharing: while it fosters trust and enhances perceived collaboration, it can occasionally hinder objective performance gains.

Uri Menkes, Assaf Hallak, Ofra Amir

Consider a setting where a pre-trained agent is operating in an environment and a human operator can decide to temporarily terminate its operation and take-over for some duration of time. These kind of scenarios are common in human-machine interactions, for example in autonomous driving, factory automation and healthcare. In these settings, we typically observe a trade-off between two extreme cases -- if no take-overs are allowed, then the agent might employ a sub-optimal, possibly dangerous policy. Alternatively, if there are too many take-overs, then the human has no confidence in the agent, greatly limiting its usefulness. In this paper, we formalize this setup and propose an explainability scheme to help optimize the number of human interventions.

Sahar Admoni, Assaf Hallak, Yftah Ziser et al.

Policies generated by Reinforcement Learning (RL) algorithms are difficult to explain to users, as they emerge from the interaction of complex reward structures and neural network representations. Consequently, analyzing and predicting agent behavior can be challenging, undermining user trust in real-world applications. To facilitate user understanding, current methods for global policy summarization typically rely on videos that demonstrate agent behavior in a subset of world states. However, users can only watch a limited number of demonstrations, constraining their understanding. Moreover, these methods place the burden of interpretation on users by presenting raw behaviors rather than synthesizing them into coherent patterns. To resolve these issues, we introduce SySLLM (Synthesized Summary using Large Language Models), advocating for a new paradigm of abstractive-textual policy explanations. By leveraging Large Language Models (LLMs)-which possess extensive world knowledge and pattern synthesis capabilities-SySLLM generates textual summaries that provide structured and comprehensible explanations of agent policies. SySLLM demonstrates that LLMs can interpret spatio-temporally structured descriptions of state-action trajectories from an RL agent and generate valuable policy insights in a zero-shot setting, without any prior knowledge or fine-tuning. Our evaluation shows that SySLLM captures key insights, such as goal preferences and exploration strategies, that were also identified by human experts. Furthermore, in a large-scale user study (with 200 participants), SySLLM summaries were preferred over demonstration-based summaries (HIGHLIGHTS) by a clear majority (75.5%) of participants.

Yotam Amitai, Ofra Amir

With Artificial Intelligence on the rise, human interaction with autonomous agents becomes more frequent. Effective human-agent collaboration requires users to understand the agent's behavior, as failing to do so may cause reduced productivity, misuse or frustration. Agent strategy summarization methods are used to describe the strategy of an agent to its destined user through demonstration. A summary's objective is to maximize the user's understanding of the agent's aptitude by showcasing its behaviour in a selected set of world states. While shown to be useful, we show that current methods are limited when tasked with comparing between agents, as each summary is independently generated for a specific agent. In this paper, we propose a novel method for generating dependent and contrastive summaries that emphasize the differences between agent policies by identifying states in which the agents disagree on the best course of action. We conduct user studies to assess the usefulness of disagreement-based summaries for identifying superior agents and conveying agent differences. Results show disagreement-based summaries lead to improved user performance compared to summaries generated using HIGHLIGHTS, a strategy summarization algorithm which generates summaries for each agent independently.

Roee Shraga, Ofra Amir, Avigdor Gal

Matching is a task at the heart of any data integration process, aimed at identifying correspondences among data elements. Matching problems were traditionally solved in a semi-automatic manner, with correspondences being generated by matching algorithms and outcomes subsequently validated by human experts. Human-in-the-loop data integration has been recently challenged by the introduction of big data and recent studies have analyzed obstacles to effective human matching and validation. In this work we characterize human matching experts, those humans whose proposed correspondences can mostly be trusted to be valid. We provide a novel framework for characterizing matching experts that, accompanied with a novel set of features, can be used to identify reliable and valuable human experts. We demonstrate the usefulness of our approach using an extensive empirical evaluation. In particular, we show that our approach can improve matching results by filtering out inexpert matchers.

Tobias Huber, Katharina Weitz, Elisabeth André et al.

With advances in reinforcement learning (RL), agents are now being developed in high-stakes application domains such as healthcare and transportation. Explaining the behavior of these agents is challenging, as the environments in which they act have large state spaces, and their decision-making can be affected by delayed rewards, making it difficult to analyze their behavior. To address this problem, several approaches have been developed. Some approaches attempt to convey the $\textit{global}$ behavior of the agent, describing the actions it takes in different states. Other approaches devised $\textit{local}$ explanations which provide information regarding the agent's decision-making in a particular state. In this paper, we combine global and local explanation methods, and evaluate their joint and separate contributions, providing (to the best of our knowledge) the first user study of combined local and global explanations for RL agents. Specifically, we augment strategy summaries that extract important trajectories of states from simulations of the agent with saliency maps which show what information the agent attends to. Our results show that the choice of what states to include in the summary (global information) strongly affects people's understanding of agents: participants shown summaries that included important states significantly outperformed participants who were presented with agent behavior in a randomly set of chosen world-states. We find mixed results with respect to augmenting demonstrations with saliency maps (local information), as the addition of saliency maps did not significantly improve performance in most cases. However, we do find some evidence that saliency maps can help users better understand what information the agent relies on in its decision making, suggesting avenues for future work that can further improve explanations of RL agents.

Isaac Lage, Daphna Lifschitz, Finale Doshi-Velez et al.

AI agents are being developed to support high stakes decision-making processes from driving cars to prescribing drugs, making it increasingly important for human users to understand their behavior. Policy summarization methods aim to convey strengths and weaknesses of such agents by demonstrating their behavior in a subset of informative states. Some policy summarization methods extract a summary that optimizes the ability to reconstruct the agent's policy under the assumption that users will deploy inverse reinforcement learning. In this paper, we explore the use of different models for extracting summaries. We introduce an imitation learning-based approach to policy summarization; we demonstrate through computational simulations that a mismatch between the model used to extract a summary and the model used to reconstruct the policy results in worse reconstruction quality; and we demonstrate through a human-subject study that people use different models to reconstruct policies in different contexts, and that matching the summary extraction model to these can improve performance. Together, our results suggest that it is important to carefully consider user models in policy summarization.

Reshef Meir, Ofra Amir, Omer Ben-Porat et al.

Truth discovery is a general name for a broad range of statistical methods aimed to extract the correct answers to questions, based on multiple answers coming from noisy sources. For example, workers in a crowdsourcing platform. In this paper, we consider an extremely simple heuristic for estimating workers' competence using average proximity to other workers. We prove that this estimates well the actual competence level and enables separating high and low quality workers in a wide spectrum of domains and statistical models. Under Gaussian noise, this simple estimate is the unique solution to the MLE with a constant regularization factor. Finally, weighing workers according to their average proximity in a crowdsourcing setting, results in substantial improvement over unweighted aggregation and other truth discovery algorithms in practice.

Gal Cohensius, Omer Ben Porat, Reshef Meir et al.

Crowdsourcing platforms offer a way to label data by aggregating answers of multiple unqualified workers. We introduce a \textit{simple} and \textit{budget efficient} crowdsourcing method named Proxy Crowdsourcing (PCS). PCS collects answers from two sets of workers: \textit{leaders} (a.k.a proxies) and \textit{followers}. Each leader completely answers the survey while each follower answers only a small subset of it. We then weigh every leader according to the number of followers to which his answer are closest, and aggregate the answers of the leaders using any standard aggregation method (e.g., Plurality for categorical labels or Mean for continuous labels). We compare empirically the performance of PCS to unweighted aggregation, keeping the total number of questions (the budget) fixed. We show that PCS improves the accuracy of aggregated answers across several datasets, both with categorical and continuous labels. Overall, our suggested method improves accuracy while being simple and easy to implement.