arXiv subjects

Reza Habibi

Publications and source records attributed to Reza Habibi.

5 recordsLinked to original sources

arXiv2026

Symbolic Mechanistic Data Attribution: Tracing Training Influence to Learned Behavioral Policies

While existing data attribution methods can identify which training examples build specific mechanistic circuits, they cannot explain how training data shapes the high-level behavioral decisions a model learns to make. To bridge this gap, we introduce Symbolic Mechanistic Data Attribution (SMDA), a framework that attributes training pairs to the interpretable symbolic policies governing model behavior. SMDA fits a closed-form Ridge regression over sparse autoencoder (SAE) features to model a target behavior, then analytically decomposes how each supervised fine-tuning example shifts that policy through feature-activation Delta_X and output-probability Delta_Y pathways. We distill a symbolic policy for refusal behavior in Llama-3.2-3B-Instruct and analyze 200 SFT training pairs. Our analysis reveals that (1) the symbolic policy's coefficients expose systematic gaps in the base model's safety behavior for categories like religious stereotyping; (2) per-feature Delta_X/Delta_Y decomposition can mechanistically explain why harmful and harmless pairs exert qualitatively different influences on certain features; and (3) individual training pairs routinely exhibit cross-feature interference, allowing SMDA to identify training pairs whose dominant effect falls on unintended features. These results demonstrate that combining mechanistic interpretability with data attribution yields a diagnostic tool that is both more fine-grained than black-box influence functions and more scalable than manual circuit analysis.

cs.LG↗

arXiv2026

Beyond Accuracy: Introducing a Symbolic-Mechanistic Approach to Interpretable Evaluation

Accuracy-based evaluation cannot reliably distinguish genuine generalization from shortcuts like memorization, leakage, or brittle heuristics, especially in small-data regimes. In this position paper, we argue for mechanism-aware evaluation that combines task-relevant symbolic rules with mechanistic interpretability, yielding algorithmic pass/fail scores that show exactly where models generalize versus exploit patterns. We demonstrate this on NL-to-SQL by training two identical architectures under different conditions: one without schema information (forcing memorization), one with schema (enabling grounding). Standard evaluation shows the memorization model achieves 94% field-name accuracy on unseen data, falsely suggesting competence. Our symbolic-mechanistic evaluation reveals this model violates core schema generalization rules, a failure invisible to accuracy metrics.

cs.LG↗

arXiv2025

What Do You Mean? Exploring How Humans and AI Interact with Symbols and Meanings in Their Interactions

Meaningful human-AI collaboration requires more than processing language; it demands a deeper understanding of symbols and their socially constructed meanings. While humans naturally interpret symbols through social interaction, AI systems often miss the dynamic interpretations that emerge in conversation. Drawing on Symbolic Interactionism theory, we conducted two studies to investigate how humans and AI co-construct symbols and their meanings. Findings provide empirical insights into how humans and conversational AI agents collaboratively shape meanings during interaction. We show how participants shift their initial definitions of meaning in response to the symbols and interpretations suggested by the conversational AI agents, especially when social context is introduced. We also observe how participants project their personal and social values into these interactions, refining meanings over time. These findings reveal that shared understanding does not emerge from mere agreement but from the bi-directional exchange and reinterpretation of symbols, suggesting new paradigms for human-AI interaction design.

cs.AI↗

arXiv2023

Modeling Player Personality Factors from In-Game Behavior and Affective Expression

Developing a thorough understanding of the target audience (and/or single individuals) is a key factor for success - which is exceptionally important and powerful for the domain of video games that can not only benefit from informed decision making during development, but ideally even tailor game content, difficulty and player experience while playing. The granular assessment of individual personality and differences across players is a particularly difficult endeavor, given the highly variant human nature, disagreement in psychological background models and because of the effortful data collection that most often builds upon long, time-consuming and deterrent questionnaires. In this work, we explore possibilities to predict a series of player personality questionnaire metrics from recorded in-game behavior and extend related work by explicitly adding affective dialog decisions to the game environment which could elevate the model's accuracy. Using random forest regression, we predicted a wide variety of personality metrics from seven established questionnaires across 62 players over 60 minute gameplay of a customized version of the role-playing game Fallout: New Vegas. While some personality variables could already be identified from reasonable underlying in-game actions and affective expressions, we did not find ways to predict others or encountered questionable correlations that could not be justified by theoretical background literature. Yet, building on the initial opportunities of this explorative study, we are striving to massively enlarge our data set to players from an ecologically valid industrial game environment and investigate the performance of more sophisticated machine learning approaches.

cs.AI↗

arXiv2023

Empathetic AI for Empowering Resilience in Games

Failure and resilience are important aspects of gameplay. This is especially important for serious and competitive games, where players need to adapt and cope with failure frequently. In such situations, emotion regulation -- the active process of modulating ones' emotions to cope and adapt to challenging situations -- becomes essential. It is one of the prominent aspects of human intelligence and promotes mental health and well-being. While there has been work on developing artificial emotional regulation assistants to help users cope with emotion regulation in the field of Intelligent Tutoring systems, little is done to incorporate such systems or ideas into (serious) video games. In this paper, we introduce a data-driven 6-phase approach to establish empathetic artificial intelligence (EAI), which operates on raw chat log data to detect key affective states, identify common sequences and emotion regulation strategies and generalizes these to make them applicable for intervention systems.

cs.HC↗