论文检索

Sridhar Mahadevan

We propose a new theoretical foundation for artificial intelligence (AI) and machine learning (ML), building on ideas in pure mathematics relating to categories and functors. This paper builds on our AAAI 2025 tutorial Thinking with Functors: Category Theory for A(G)I, which provides background material. In addition, our recent papers on intuitionistic j-do calculus in Topos Causal Models} and GAIA: Categorical Foundations of Generative AI, illustrate how to generalize well-known formalisms in AI, such as causal inference and deep learning, to a category-theoretic setting.

Khadija Zanna, Akane Sano

Ensuring fairness in machine learning requires understanding how sensitive attributes like race or gender causally influence outcomes. Existing causal discovery (CD) methods often struggle to recover fairness-relevant pathways in the presence of noise, confounding, or data corruption. Large language models (LLMs) offer a complementary signal by leveraging semantic priors from variable metadata. We propose a hybrid LLM-guided CD framework that extends a breadth-first search strategy with active learning and dynamic scoring. Variable pairs are prioritized for querying using a composite score combining mutual information, partial correlation, and LLM confidence, enabling more efficient and robust structure discovery. To evaluate fairness sensitivity, we introduce a semi-synthetic benchmark based on the UCI Adult dataset, embedding domain-informed bias pathways alongside noise and latent confounders. We assess how well CD methods recover both global graph structure and fairness-critical paths (e.g., sex→education→income). Our results demonstrate that LLM-guided methods, including our active, dynamically scored variant, outperform baselines in recovering fairness-relevant structure under noisy conditions. We analyze when LLM-driven insights complement statistical dependencies and discuss implications for fairness auditing in high-stakes domains.

Yingying Wang, Qin Ni, Haoxin Xu, Jiaqi Yin, Tingjiang Wei

Artificial intelligence is playing an increasingly important role in supporting decision-making, particularly in educational contexts, where it serves as a critical tool to assist teacher judgment and optimize instructional decisions. However, limited research has examined how different AI-assisted decision-making paradigms influence the Performance of human-AI collaboration, as well as the underlying psychological mechanisms and causal pathways. Therefore, this study investigated 59 pre-service teachers to examine how AI-assisted decision-making paradigms and human-AI consistency influenced their psychological states and task performance. Specifically, this study employed a two-factor mixed experimental design, with the AI-assisted decision-making paradigms as the between-subjects factor and human-AI consistency as the within-subjects factor. Data were analyzed using the Bayesian cumulative link mixed model and structural equation modeling. The results reveal that AI-assisted decision-making paradigms do not have a significant direct effect on task performance. However, when the moderating role of human-AI decision consistency is taken into account, the effect of AI-assisted decision-making paradigms on task performance can exert its influence indirectly through a sequential psychological pathway involving users’ confidence and their trust in the AI. Consistency between human and AI decisions not only significantly enhances users’ trust in AI, confidence, and task performance, but the proportion of consistent decisions also significantly moderates the impact of AI-assisted decision-making paradigms on users’ confidence levels. Notably, our findings indicate that users maintain a moderately level of trust in AI even when their decisions diverge from those of AI. In summary, this study highlights the mediating mechanism by which AI-assisted decision-making paradigms influence task performance through psychological states and identifies the moderating role of human-AI consistency in this pathway. These findings advance the theoretical understanding of human-AI interaction models in educational contexts and offer mechanistic insights to guide the optimization of instructional AI systems.

Markus B. Pettersson, Connor Thomas Jerzak, Adel Daoud

Machine learning models trained on Earth observation data, such as satellite imagery, have demonstrated significant promise in predicting household-level wealth indices, enabling the creation of high-resolution wealth maps that can be leveraged across multiple causal trials while addressing chronic data scarcity in global development research. However, because standard training objectives prioritize overall predictive accuracy, these predictions often suffer from shrinkage toward the mean, leading to attenuated estimates of causal treatment effects and limiting their utility in policy evaluations. Existing debiasing methods, such as Prediction-Powered Inference (PPI), can handle this attenuation bias but require additional fresh ground-truth data at the downstream stage of causal inference, which restricts their applicability in data-scarce environments. We introduce and evaluate two post-hoc correction methods—Linear Calibration Correction (LCC) and a Tweedie's correction approach—that substantially reduce shrinkage-induced prediction bias without relying on newly collected labeled data. LCC applies a simple linear transformation estimated on a held-out calibration split; Tweedie's method locally de-shrink predictions using density score estimates and a noise scale learned upstream. We provide practical diagnostics for when a correction is warranted and discuss practical limitations. Across analytical results, simulations, and experiments with Demographic and Health Surveys (DHS) data, both approaches reduce attenuation; Tweedie's correction yields nearly unbiased treatment-effect estimates, enabling a "one map, many trials" paradigm. Although we demonstrate on EO-ML wealth mapping, the methods are not geospatial-specific: they apply to any setting where imputed outcomes are reused downstream (e.g., pollution indices, population density, or LLM-derived indicators).

Ayan Majumdar, Deborah D. Kanubala, Kavya Gupta, Isabel Valera

Fairness studies of algorithmic decision-making systems often simplify complex decision processes, such as bail or lending decisions, into binary classification tasks (e.g., approve or not approve). However, these approaches overlook that such decisions are not inherently binary; they also involve non-binary treatment decisions (e.g., loan or bail terms) that can influence the downstream outcomes (e.g., loan repayment or reoffending). We argue that treatment decisions are integral to the decision-making process and, therefore, should be central to fairness analyses. Consequently, we propose a causal framework that extends and complements existing fairness notions by explicitly distinguishing between decision-subjects’ covariates and the treatment decisions. Our framework leverages path-specific counterfactual reasoning to: (i) measure treatment disparity and its downstream effects in historical data; and (ii) mitigate the impact of past unfair treatment decisions when automating decision-making. We use our framework to empirically analyze four widely used loan approval datasets to reveal potential disparity in non-binary treatment decisions and their discriminatory impact on outcomes, highlighting the need to incorporate treatment decisions in fairness assessments. Finally, by intervening in treatment decisions, we show that our framework effectively mitigates treatment discrimination from historical loan approval data to ensure fair risk score estimation and (non-binary) decision-making processes that benefit all stakeholders.

Woojin Kim, Changkwon Lee, Hyeoncheol Kim

Using Artificial Intelligence to improve teaching and learning benefits greater adaptivity and scalability in education. Knowledge Tracing (KT) is recognized for student modeling task due to its superior performance and application potential in education. To this end, we conceptualize and investigate counterfactual explanation as the connection from XAI for KT to education. Counterfactual explanations offer actionable recourse, are inherently causal and local, and easy for educational stakeholders to understand who are often non-experts. We propose KTCF, a counterfactual explanation generation method for KT that accounts for knowledge concept relationships, and a post-processing scheme that converts a counterfactual explanation into a sequence of educational instructions. We experiment on a large-scale educational dataset and show our KTCF method achieves superior and robust performance over existing methods, with improvements ranging from 5.7% to 34% across metrics. Additionally, we provide a qualitative evaluation of our post-processing scheme, demonstrating that the resulting educational instructions help in reducing large study burden. We show that counterfactuals have the potential to advance the responsible and practical use of AI in education. Future works on XAI for KT may benefit from educationally grounded conceptualization and developing stakeholder-centered methods.

Qinghong Guo, Yu Wang, Ji Cao, Tongya Zheng, Junshu Dai, Bingde Hu, Shunyu Liu, Canghong Jin

Road network representation learning (RNRL) has attracted increasing attention from both researchers and practitioners as various spatiotemporal tasks are emerging. Recent advanced methods leverage Graph Neural Networks (GNNs) and contrastive learning to characterize the spatial structure of road segments in a self-supervised paradigm. However, spatial heterogeneity and temporal dynamics of road networks raise severe challenges to the neighborhood smoothing mechanism of self-supervised GNNs. To address these issues, we propose a Dual-branch Spatial-Temporal self-supervised representation framework for enhanced road representations, termed as DST. On one hand, DST designs a mix-hop transition matrix for graph convolution to incorporate dynamic relations of roads from trajectories. Besides, DST contrasts road representations of the vanilla road network against that of hypergraphs in a spatial self-supervised way. The hypergraph is newly built based on three types of hyperedges to capture long-range relations. On the other hand, DST performs next token prediction as the temporal self-supervised task on the sequences of traffic dynamics based on a causal Transformer, which is further regularized by differentiating traffic modes of weekdays from those of weekends. Extensive experiments against state-of-the-art methods verify the superiority of our proposed framework. Moreover, the comprehensive spatiotemporal modeling facilitates DST to excel in zero-shot learning scenarios.

Guangrui Fan, Dandan Liu, Lihu Pan

Online peer-support communities are vital for mental health, but their therapeutic benefit hinges on receiving a timely and helpful first reply. Posts that languish unanswered can exacerbate feelings of distress and abandonment. This paper develops and validates an integrated framework to predict, explain, and reduce this ``reply gap" on Reddit. First, using survival analysis on over 91,000 posts (2018–2025), we show that a deep learning model (DySurv) can accurately predict reply times (C-Index = 0.742), with a post's lexico-semantic content being a far stronger predictor than author history. Second, moving from correlation to causation, we use a causal inference framework on 48,612 posts to estimate the effect of different support types. We find that initial replies providing emotional support are most effective, increasing the odds of a positive user response by 49% (OR=1.49), an effect most pronounced for high-risk users. Third, we operationalize these insights in RiskMatch, a recommender system that routes at-risk posts to historically effective helpers. Rigorous counterfactual evaluation using inverse propensity scoring (IPS)—a method that corrects for biases in historical data—demonstrates that our system reduces the median wait time by 26 minutes for the highest-risk quintile. This work provides a validated, data-driven methodology to build more responsive and effective peer-support ecosystems, offering a concrete pathway to ensure fewer calls for help go unanswered.

Nirmalendu Prakash, Yeo Wei Jie, Amir Abdullah, Ranjan Satapathy, Erik Cambria, Roy Ka-Wei Lee

Refusal on harmful prompts is a key safety behaviour in instruction‑tuned large language models (LLMs), yet the internal causes of this behaviour remain poorly understood. We study two public instruction tuned models—Gemma‑2-2B‑IT and LLaMA‑3.1-8B‑IT using sparse autoencoders (SAEs) trained on residual‑stream activations. Given a harmful prompt, we search the SAE latent space for feature sets whose ablation flips the model from refusal to compliance, demonstrating causal influence and creating a jailbreak. Our search proceeds in three stages: 1. Refusal Direction - Finding a refusal mediating direction and collecting SAE features close to that direction, followed by 2. Greedy Filtering - to prune this set to obtain a minimal set and finally 3. Interaction Discovery - a factorization‑machine (FM) model that captures non‑linear interactions among the remaining active features and the minimal set. This pipeline yields a broad set of jailbreak-critical features, offering insight into the mechanistic basis of refusal. Moreover, we also find evidence of redundant features which remain dormant unless earlier features are suppressed. Our findings highlight the potential for fine-grained auditing and targeted intervention in safety behaviours by manipulating the interpretable latent space.

Prudhviraj Naidu, Zixian Wang, Leon Bergen, Ramamohan Paturi

We train Transformer-based language models on ten foundational algorithmic tasks and observe pronounced phase transitions in their loss curves that deviate from established power-law scaling trends. Over large ranges of compute, the validation loss barely improves, then abruptly decreases. Probing the models’ internal representations reveals that quiet features are learned prior to any decrease in task loss. These quiet features represent intermediate algorithmic computations that do not by themselves improve the output loss. Ablation experiments demonstrate that individual quiet features are causally necessary for task performance. Our results demonstrate that substantial representational progress can remain hidden beneath an apparently flat loss curve, challenging the prevailing use of cross‑entropy as a proxy for learning and motivating richer diagnostics for monitoring model training.

Jie Lin, Lei Jiang, Zongyi Chen, Liansheng Wang

With the advancement of information retrieval (IR) technologies toward deep semantic understanding, reasoning-based methods—featuring explicit chain-of-thought generation—have demonstrated significant advantages in multi-hop and causal reasoning tasks. However, in complex clinical case retrieval scenarios, implicit reasoning cues within clinical data often hinder current models from effectively capturing deep semantic associations between queries and cases. Query rewriting and expansion techniques based on reasoning offer a promising solution to this challenge by uncovering and completing the latent clinical intent behind user queries, thereby enhancing semantic coverage and reasoning sensitivity. In this paper, we propose CRAF, a clinically adaptive reasoning framework tailored for similar case retrieval. Our method generates clinical reasoning paths and incorporates a fine-grained semantic reward mechanism, enabling efficient query rewriting through reinforcement learning. Experimental results on the PMC-Patients benchmark demonstrate that CRAF consistently delivers robust improvements across multiple retrieval tasks, achieving reasoning performance comparable to that of commercial models.

Zidong Wang, Xi Lin, Chuchao He, Xiaoguang Gao

Robust causal discovery from observational data under imperfect prior knowledge remains a significant and largely unresolved challenge. Existing methods typically presuppose perfect priors or can only handle specific, pre-identified error types. And their performance degrades substantially when confronted with flawed constraints of unknown location and type. This decline arises because most of them rely on inflexible and biased thresholding strategies that may conflict with the data distribution. To overcome these limitations, we propose to harmonizes knowledge and data through prior alignment and conflict resolution. First, we assess the credibility of imperfect structural constraints through a surrogate model, which then guides a sparse penalization term measuring the loss between the learned and constrained adjacency matrices. We theoretically prove that, under ideal assumption, the knowledge-driven objective aligns with the data-driven objective. Furthermore, to resolve conflicts when this assumption is violated, we introduce a multi-task learning framework optimized via multi-gradient descent, jointly minimizing both objectives. Our proposed method is robust to both linear and nonlinear settings. Extensive experiments, conducted under diverse noise conditions and structural equation model types, demonstrate the effectiveness and efficiency of our method under imperfect structural constraints.

Pablo Torrijos, Jose M. Puerta, Juan A. Aledo, José A. Gámez

This paper presents the Min-Cut Bayesian Network Consensus (MCBNC) algorithm, a greedy method for structural consensus of Bayesian Networks (BNs), with applications in federated learning and model aggregation. MCBNC prunes weak edges from an initial unrestricted fusion using a structural score based on min-cut analysis, integrated into a modified Backward Equivalence Search (BES) phase of the Greedy Equivalence Search (GES) algorithm. The score quantifies edge support across input networks and is computed using max-flow. Unlike methods with fixed treewidth bounds, MCBNC introduces a pruning threshold θ that can be selected post hoc using only structural information. Experiments on real-world BNs show that MCBNC yields sparser, more accurate consensus structures than both canonical fusion and the input networks. The method is scalable, data-agnostic, and well-suited for distributed or federated structural learning of BNs or causal discovery.

Nicholas Tagliapietra, Katharina Ensinger, Christoph Zimmer, Osman Mian

Real world systems evolve in continuous-time according to their underlying causal relationships, yet their dynamics are often unknown. Existing approaches to learning such dynamics typically either discretize time ---leading to poor performance on irregularly sampled data--- or ignore the underlying causality. We propose CADYT, a novel method for causal discovery on dynamical systems addressing both these challenges. In contrast to state-of-the-art causal discovery methods that model the problem using discrete-time Dynamic Bayesian networks, our formulation is grounded in Difference-based causal models, which allow milder assumptions for modeling the continuous nature of the system. CADYT leverages exact Gaussian Process inference for modeling the continuous-time dynamics which is more aligned with the underlying dynamical process. We propose a practical instantiation that identifies the causal structure via a greedy search guided by the Algorithmic Markov Condition and Minimum Description Length principle. Our experiments show that CADYT outperforms state-of-the-art methods on both regularly and irregularly-sampled data, discovering causal networks closer to the true underlying dynamics.

Kohsuke Kubota, Shonosuke Sugasawa

Many marketing applications, including credit card incentive programs, offer rewards to customers who exceed specific spending thresholds to encourage increased consumption. Quantifying the causal effect of these thresholds on customers is crucial for effective marketing strategy design. Although regression discontinuity design is a standard method for such causal inference tasks, its assumptions can be violated when customers, aware of the thresholds, strategically manipulate their spending to qualify for the rewards. To address this issue, we propose a novel framework for estimating the causal effect under threshold manipulation. The main idea is to model the observed spending distribution as a mixture of two distributions: one representing customers strategically affected by the threshold, and the other representing those unaffected. To fit the mixture model, we adopt a two-step Bayesian approach consisting of modeling non-bunching customers and fitting a mixture model to a sample around the threshold. We show posterior contraction of the resulting posterior distribution of the causal effect under large samples. Furthermore, we extend this framework to a hierarchical Bayesian setting to estimate heterogeneous causal effects across customer subgroups, allowing for stable inference even with small subgroup sample sizes. We demonstrate the effectiveness of our proposed methods through simulation studies and illustrate their practical implications using a real-world marketing dataset.

Yuta Kawakami, Jin Tian

Counterfactual decision-making in the face of uncertainty involves selecting the optimal action from several alternatives using causal reasoning. Decision-makers often rank expected potential outcomes (or their corresponding utility and desirability) to compare the preferences of candidate actions. In this paper, we study new counterfactual decision-making rules by introducing two new metrics: the probabilities of potential outcome ranking (PoR) and the probability of achieving the best potential outcome (PoB). PoR reveals the most probable ranking of potential outcomes for an individual, and PoB indicates the action most likely to yield the top-ranked outcome for an individual. We then establish identification theorems and derive bounds for these metrics, and present estimation methods. Finally, we perform numerical experiments to illustrate the finite-sample properties of the estimators and demonstrate their application to a real-world dataset.

Johan De Aguas

We introduce two families of stochastic interventions with discrete treatments that connect causal modeling to cost-sensitive decision making. The interventions arise from a cost-penalized information projection of the independent product of the organic propensity scores and a reference policy, yielding closed-form Boltzmann-Gibbs couplings. The induced marginals define modified stochastic policies that interpolate smoothly, via a tilt parameter, from the organic law or from the reference law toward a product-of-experts limit when all destination costs are strictly positive. The first family recovers and extends incremental propensity score interventions, retaining identification without global positivity. For inference on the expected outcomes after these policies, we derive the efficient influence functions under a nonparametric model and construct one-step estimators. In simulations, the proposed estimators improve stability and robustness to nuisance misspecification relative to plug-in baselines. The framework can operationalize graded scientific hypotheses under realistic constraints. Because inputs are modular, analysts can sweep feasible policy spaces, prototype candidates, and align interventions with budgets and logistics before committing experimental resources.

Harrison Oates, Pascal Bercher

Modern planning systems utilize various plan representations - sequential, parallel, partially ordered (PO), and partial-order causal link (POCL) - each with different models for concurrency. These formalisms are often implicitly assumed to have the same base properties, particularly regarding makespan. We challenge this assumption, proving the relationship between them is fundamentally asymmetric. Our analysis shows conversions from plans with rigid concurrency layers (sequential, parallel) to those with flexible partial orders (PO, POCL) can preserve makespan. However, the reverse generally fails; the flexible orderings in PO/POCL plans can yield shorter makespans for solutions that cannot be represented in parallel plans without serialization. We prove that finding an optimal parallel representation for a given POCL plan is NP-complete, resolving a key question about their practical interchangeability. We also provide tight complexity bounds for makespan-bounded plan existence. Notably, our results disprove a claim in the literature that planning graph-based planners maximize concurrency by minimizing the critical path in derived PO plans.

Liu Yu, Zhonghao Chen, Ping Kuang, Zhikun Feng, Fan Zhou, Lan Wang, Gillian Dobbie

Object hallucination remains a critical challenge in Large Vision-Language Models (LVLMs), where models generate content inconsistent with visual inputs. Existing language-decoder based mitigation approaches often regulate visual or textual attention independently, overlooking their interaction as two key causal factors. To address this, we propose Owl (Bi-mOdal attention reWeighting for Layer-wise hallucination mitigation), a causally-grounded framework that models hallucination process via a structural causal graph, treating decomposed visual and textual attentions as mediators. We introduce VTACR (Visual-to-Textual Attention Contribution Ratio), a novel metric that quantifies the modality contribution imbalance during decoding. Our analysis reveals that hallucinations frequently occur in low-VTACR scenarios, where textual priors dominate and visual grounding is weakened. To mitigate this, we design a fine-grained attention intervention mechanism that dynamically adjusts token- and layer-wise attention guided by VTACR signals. Finally, we propose a dual-path contrastive decoding strategy: one path emphasizes visually grounded predictions, while the other amplifies hallucinated ones -- letting visual truth shine and hallucination collapse. Experimental results on the POPE and CHAIR benchmarks show that Owl achieves significant hallucination reduction, setting a new SOTA in faithfulness while preserving vision-language understanding capability. Our code is available at https://github.com/CikZ2023/OWL

Yiran Yang, Zhaowei Liu, Yuan Yuan, Yukun Song, Xiong Ma, Yinghao Song, Xiangji Zeng, Lu Sun, Yulu Wang, Hai Zhou 等

Short-video platforms now host vast multimodal ads whose deceptive visuals, speech and subtitles demand finer-grained, policy-driven moderation than community safety filters. We present BLM-Guard, a content-audit framework for commercial ads that fuses Chain-of-Thought reasoning with rule-based policy principles and a critic-guided reward. A rule-driven ICoT data-synthesis pipeline jump-starts training by generating structured scene descriptions, reasoning chains and labels, cutting annotation costs. Reinforcement learning then refines the model using a composite reward balancing causal coherence with policy adherence. A multitask architecture models intra-modal manipulations (e.g., exaggerated imagery) and cross-modal mismatches (e.g., subtitle–speech drift), boosting robustness. Experiments on real short-video ads show BLM-Guard surpasses strong baselines in accuracy, consistency and generalization.