AI Biases as Asymmetries: A Review to Guide Practice Gabriella Waters (CEAMLS, Morgan State University)* Phillip Honenberger (CEAMLS, Morgan State University)* *Equal contribution [Preprint - Nov. 21, 2024] Abstract The understanding of bias in AI is currently undergoing a revolution. Initially understood as errors or flaws, biases are increasingly recognized as integral to AI systems and sometimes preferable to less biased alternatives. In this paper we review the reasons for this changed understanding and provide new guidance on two questions: First, how should we think about and measure biases in AI systems, consistent with the new understanding? Second, what kinds of bias in an AI system should we accept or even amplify, and what kinds should we minimize or eliminate, and why? The key to answering both questions, we argue, is to understand biases as "violations of a symmetry standard" (following Kelly). We distinguish three main types of asymmetry in AI systems - error biases, inequality biases, and process biases - and highlight places in the pipeline of AI development and application where bias of each type is likely to be good, bad, or inevitable. Introduction The understanding of bias in AI is currently undergoing a revolution. Initially perceived as errors or flaws, biases are increasingly recognized as integral to AI systems and sometimes preferable to less biased alternatives. Cognitive psychology and statistics have informed this shift, highlighting the benefits and costs of biases in decision-making processes. Cognitive psychology presents biases as often helpful in making decisions under conditions of uncertainty. Similarly, statistical methods acknowledge biases as often useful and sometimes necessary for making inferences from data. These insights have been instrumental in redefining biases as not inherently negative, but as sometimes essential components that can and should be harnessed to improve AI systems.

Cognitive biases, systematic deviations from rationality in judgment, pose significant challenges in generating objective content. This paper introduces a novel approach for real-time cognitive bias detection in user-generated text using large language models (LLMs) and advanced prompt engineering techniques. The proposed system analyzes textual data to identify common cognitive biases such as confirmation bias, circular reasoning, and hidden assumption. By designing tailored prompts, the system effectively leverages LLMs' capabilities to both recognize and mitigate these biases, improving the quality of human-generated content (e.g., news, media, reports). Experimental results demonstrate the high accuracy of our approach in identifying cognitive biases, offering a valuable tool for enhancing content objectivity and reducing the risks of biased decisionmaking. Introduction Cognitive biases are systematic patterns of deviation from rational judgment, affecting decision-making processes across various domains, including media, policy-making, and legal reasoning. With the rapid expansion of artificial intelligence (AI) applications, large language models (LLMs) have demonstrated significant potential in processing and evaluating vast amounts of textual information. However, existing research has largely focused on mitigating biases within AI-generated outputs rather than leveraging AI to detect biases in human-generated content. This gap presents a critical challenge in ensuring transparency and fairness in AI-assisted decision-making. This study explores the application of structured prompt engineering as a novel approach to improving LLM accuracy in detecting cognitive biases.

Learning therapeutic counseling involves significant role-play experience with mock patients, with current manual training methods providing only intermittent granular feedback. We seek to accelerate and optimize counselor training by providing frequent, detailed feedback to trainees as they interact with a simulated patient. Our first application domain involves training motivational interviewing skills for counselors. Motivational interviewing is a collaborative counseling style in which patients are guided to talk about changing their behavior, with empathetic counseling an essential ingredient. We developed and evaluated an LLM-powered training system that features a simulated patient and visualizations of turn-by-turn performance feedback tailored to the needs of counselors learning motivational interviewing. We conducted an evaluation study with professional and student counselors, demonstrating high usability and satisfaction with the system. We present design implications for the development of automated systems that train users in counseling skills and their generalizability to other types of social skills training.

Language Models (LMs) have demonstrated impressive capabilities in solving complex reasoning tasks, particularly when prompted to generate intermediate explanations. However, it remains an open question whether these intermediate reasoning traces represent a dynamic, evolving thought process or merely reflect sophisticated pattern recognition acquired during large scale pre training. Drawing inspiration from human cognition, where reasoning unfolds incrementally as new information is assimilated and internal models are continuously updated, we propose to delve deeper into the mental model of various LMs. We propose a new way to assess the mental modeling of LMs, where they are provided with problem details gradually, allowing each new piece of data to build upon and refine the model's internal representation of the task. We systematically compare this step by step mental modeling strategy with traditional full prompt methods across both text only and vision and text modalities. Experiments on the MathWorld dataset across different model sizes and problem complexities confirm that both text-based LLMs and multimodal LMs struggle to create mental representations, questioning how their internal cognitive processes work.

Predicting the behavior of human participants in strategic settings is an important problem in many domains. Most existing work either assumes that participants are perfectly rational, or attempts to directly model each participant's cognitive processes based on insights from cognitive psychology and experimental economics. In this work, we present an alternative, a deep learning approach that automatically performs cognitive modeling without relying on such expert knowledge. We introduce a novel architecture that allows a single network to generalize across different input and output dimensions by using matrix units rather than scalar units, and show that its performance significantly outperforms that of the previous state of the art, which relies on expert-constructed features.

Instructional designers face an overwhelming array of design choices, making it challenging to identify the most effective interventions. To address this issue, I propose the concept of a Model Human Learner, a unified computational model of learning that can aid designers in evaluating candidate interventions. This paper presents the first successful demonstration of this concept, showing that a computational model can accurately predict the outcomes of two human A/B experiments -- one testing a problem sequencing intervention and the other testing an item design intervention. It also demonstrates that such a model can generate learning curves without requiring human data and provide theoretical insights into why an instructional intervention is effective. These findings lay the groundwork for future Model Human Learners that integrate cognitive and learning theories to support instructional design across diverse tasks and interventions.

Recent cognitive modeling studies have reported that larger language models (LMs) exhibit a poorer fit to human reading behavior, leading to claims of their cognitive implausibility. In this paper, we revisit this argument through the lens of mechanistic interpretability and argue that prior conclusions were skewed by an exclusive focus on the final layers of LMs. Our analysis reveals that next-word probabilities derived from internal layers of larger LMs align with human sentence processing data as well as, or better than, those from smaller LMs. This alignment holds consistently across behavioral (self-paced reading times, gaze durations, MAZE task processing times) and neurophysiological (N400 brain potentials) measures, challenging earlier mixed results and suggesting that the cognitive plausibility of larger LMs has been underestimated. Furthermore, we first identify an intriguing relationship between LM layers and human measures: earlier layers correspond more closely with fast gaze durations, while later layers better align with relatively slower signals such as N400 potentials and MAZE processing times. Our work opens new avenues for interdisciplinary research at the intersection of mechanistic interpretability and cognitive modeling.

Listeners adapt language comprehension based on their mental representations of speakers, but how these representations are dynamically updated remains unclear. We investigated whether listeners probabilistically adapt their comprehension based on the likelihood of speakers producing stereotype-incongruent utterances. Our findings reveal two potential mechanisms: a speaker-general mechanism that adjusts overall expectations about speaker-content relationships, and a speaker-specific mechanism that updates individual speaker models. In two EEG experiments, participants heard speakers make stereotype-congruent or incongruent utterances, with incongruency base rate manipulated between blocks. In Experiment 1, speaker incongruency modulated both high-beta (21-30 Hz) and theta (4-6 Hz) oscillations: incongruent utterances decreased oscillatory power in low base rate condition but increased it in high base rate condition. The theta effect varied with listeners' openness trait: less open participants showed theta increases to speaker-incongruencies, suggesting maintenance of speaker-specific information, while more open participants showed theta decreases, indicating flexible model updating. In Experiment 2, we dissociated base rate from the target speaker by manipulating the overall base rate using an alternative non-target speaker. Only the high-beta effect persisted, showing power decrease for speaker-incongruencies in low base rate condition but no effect in high base rate condition. The high-beta oscillations might reflect the speaker-general adjustment, while theta oscillations may index the speaker-specific model updating. These findings provide evidence for how language processing is shaped by social cognition in real time.

Computational cognitive models, which formalize theories of cognition, enable researchers to quantify cognitive processes and arbitrate between competing theories by fitting models to behavioral data. Traditionally, these models are handcrafted, which requires significant domain knowledge, coding expertise, and time investment. Previous work has demonstrated that Large Language Models (LLMs) are adept at pattern recognition in-context, solving complex problems, and generating executable code. In this work, we leverage these abilities to explore the potential of LLMs in automating the generation of cognitive models based on behavioral data. We evaluated the LLM in two different tasks: model identification (relating data to a source model), and model generation (generating the underlying cognitive model). We performed these tasks across two cognitive domains - decision making and learning. In the case of data simulated from canonical cognitive models, we found that the LLM successfully identified and generated the ground truth model. In the case of human data, where behavioral noise and lack of knowledge of the true underlying process pose significant challenges, the LLM generated models that are identical or close to the winning model from cognitive science literature. Our findings suggest that LLMs can have a transformative impact on cognitive modeling. With this project, we aim to contribute to an ongoing effort of automating scientific discovery in cognitive science.

It's 1968, and Norm and Claudette are having lunch. Norm is explaining his position that all human languages share deep underlying structure and has worked out careful theories showing how the surface forms of language can be derived from these underlying principles. Claudette, whose favorite movie is the recently released 2001: A Space Odyssey and who particularly loves the HAL character, wants to make machines that could talk with us in any human language. Claudette asks Norm whether Norm thinks his theories could be useful for building such a system. Norm says he is interested in human language and the human mind, found HAL creepy, and isn't sure why Claudette is so interested in building chatbots or what good would come of that. Nonetheless, they both agree that it seems likely that, if Norm's theories are right (and he sure thinks they are!), they could be used to work out the fundamental rules and operations underlying human language in general--and that should, in principle, prove useful for building Claudette's linguistic machines. Claudette is very open to this possibility: all she wants is a machine that talks and understands. She doesn't really care how it happens. Norm and Claudette have very different goals, but they enjoy their conversations and are optimistic that they can both help each other.