Prediction & Perception

This research presents a novel dynamic threshold model to address the pervasive inconsistencies observed in time preference, moving beyond the limitations of traditional static models that rely on fixed parameters for present bias and discount rates. Motivated by empirical evidence demonstrating that time preferences are highly sensitive to both internal and external influences, this paper proposes a subjective threshold that is recursively updated based on a complex interplay of factors. Specifically, the threshold is dynamically adjusted according to perceived uncertainty, stochastic noise, past experiences as captured by a persistence parameter, and external shocks. The model elucidates how heightened uncertainty and stochastic noise amplify preference reversals near decision thresholds, while also demonstrating how cumulative stochastic shocks embed persistent instability into preference trajectories, hindering reversion to stable baselines. The proposed adaptive persistence mechanism allows agents to strategically recalibrate their sensitivity to volatility, thereby promoting long-term stability while maintaining responsiveness to environmental changes. The framework is supported by extensive literature review and empirical evidence demonstrating the model's capacity to capture complex intertemporal choice behavior, including cognitive processing delays, behavioral instability near uncertainty thresholds, and the enduring effects of environmental shocks. The theoretical propositions are shown to align with findings from a diverse array of studies and offer insights for policy interventions aimed at stabilizing preferences in uncertain decision-making contexts. This dynamic threshold model provides a robust foundation for future work aimed at advancing the understanding of intertemporal choices, while offering a framework for improving both theoretical rigor and practical relevance.

This is an in-person presentation on July 28, 2025 (15:00 ~ 15:20 EDT).

Using computational modeling, we systematically evaluated hypotheses about the mechanisms underlying implicit and explicit ensemble perception. Ensemble perception (EP) is the ability to summarize properties of objects in a visual array. Implicit EP tasks ask participants to judge whether a test object belongs to a studied array, with a bias toward the ensemble’s central tendency serving as a measure of ensemble representation. Explicit EP tasks ask participants to compare a test object with the ensemble mean, with performance as a measure of ensemble representation. We explored hypotheses about perceptual encoding of individual objects under capacity limitations (subsampling, noisy encoding, and/or scaled memory strength), ensemble representation (abstraction vs. exemplar), and decision processes in implicit and explicit EP tasks. Models assuming imperfect perceptual encoding and exemplar-based representations jointly accounted for the central tendency bias and poor recognition of individual objects in implicit EP, as well as set-size and variance effects in explicit EP, across a wide range of parameters. Prototype models accounted for implicit EP phenomena and set-size effects but failed to predict variance effects on accuracy. We further evaluated models’ ability to predict variance effects in choice and response time (RT). Exemplar models captured variance effects in both choice and RT, with higher ensemble variance reducing accuracy and increasing RT. In contrast, prototype models qualitatively failed to predict the variance effect on RT. Our modeling framework provides a theoretical space for testing hypotheses regarding computational mechanisms involved in ensemble perception.

This is an in-person presentation on July 28, 2025 (15:20 ~ 15:40 EDT).