The resource-rationality in heuristics applied by humans for planning in probabilistic environments
Year: 2025
Type: dissertation
Abstract: Human planning often relies on heuristics, i.e., planning strategies that selectively ignore parts of the environmental information to reduce cognitive costs. According to the framework of resource-rationality, such heuristics reflect adaptive strategies that optimize the trade-off between decision quality and cognitive costs within the limits of available cognitive and temporal resources humans have. Previous studies on resource-rational planning have primarily focused on deterministic environments. However, real-world planning typically involves uncertainty, requiring heuristics that are sensitive to probabilistic outcomes. This thesis investigates whether human planners apply resource-rational heuristics in probabilistic settings, how heuristic selection is influenced by individual cognitive constraints, and how heuristics are adapted to dynamic environmental features. To address these questions, two behavioral studies were conducted. In both studies, participants performed probabilistic three-step sequential decision-making tasks. Planning was formalized within the model-based reinforcement learning framework, in which planning heuristics are modeled as selective search through a decision tree and planning costs are quantified as decision tree complexity. Bayesian inference was used to infer latent planning parameters and analyze behavior in terms of cost-efficiency and heuristic selection. Study I assessed whether young adults (18-35 years, n = 57) and older adults (65-75 years, n = 50) reduced planning costs by focusing on the most probable branches of the decision tree while pruning, i.e. ignoring branches of lower probability (low-probability pruning). This heuristic was resource-rational in the given task environment. Additionally, the study examined whether this heuristic was additionally influenced by probability discounting - a decision bias in which outcomes are devalued as the odds against receiving them increase - potentially affecting the selection of heuristics beyond considerations of cognitive cost. Study I showed that participants of both age groups applied the resource-rational pruning heuristic and discounted probabilistic outcomes. Older adults showed stronger discounting, shallower planning depth, and increased decision noise, all of which contributed to lower planning performance compared to younger adults. The findings indicate that resource-rational heuristics are applied in probabilistic environments and that heuristic selection is influenced by age-related differences in cognitive resources and outcome evaluation. Study II assessed how participants (20-35 years, N = 71) adapt resource-rational heuristics to varying environmental features in a novel probabilistic planning task. Again, the low-probability pruning heuristic was resource-rational but also introduced potential replanning costs when previously pruned outcomes occurred. Replanning costs were systematically manipulated to assess their effect on heuristic adaptation, specifically, whether higher costs led to reduced initial planning expenditures in terms of shallower planning depth to limit overall costs. Study II replicated the use of the resource-rational low-probability pruning heuristic in a new task environment and additionally showed that participants dynamically adjusted their planning depth: higher replanning costs led to shallower initial planning, while lower costs encouraged deeper planning. This suggests that planners regulate planning expenditures in response to environmental demands by balancing immediate and future costs. The observed behavior indicates a meta-cognitive component of resource-rational heuristic selection. Together, both studies provide empirical evidence that human planners apply resource-rational heuristics and extend previous findings from deterministic to probabilistic environments. The findings suggest that planners estimate how the probabilistic structure of the environment influences expected outcomes and adapt their heuristics accordingly. The selection of heuristics is further adapted to changing environments through dynamic cost estimation and influenced by individual age-related differences. Moreover, this work refines current conceptualizations of planning costs by demonstrating that heuristic selection incorporates both immediate and anticipated future costs. In addition to its theoretical and empirical contributions, this dissertation provides a methodological foundation for studying resource-rational planning in dynamic, probabilistic environments. The newly developed task design in Study II enables the systematic manipulation of planning costs, allowing for the investigation of underlying mechanisms of the selection of heuristics. Future research should build on this foundation to assess generalizable mechanisms underlying the selection of planning heuristics, incorporate more ecologically valid environments, and further investigate how meta-cognitive processes regulate the efficient allocation of cognitive resources in human planning.
Author: Sass, Sophia-Helen
Institution: Technische Universität Dresden
Language: English
Subfield: General Decision Sciences
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
Funder: Deutsche Forschungsgemeinschaft
Awards: SFB 940: Volition und kognitive Kontrolle: Mechanismen, Modulatoren, Dysfunktionen, G3434867832, G3986097662, G4094304398, TRR 265: Verlust und Wiedererlangung der Kontrolle über den Drogenkonsum: Von Trajektorien über Mechanismen bis hin zu Interventionen, GRK 2773: Risikofaktoren und Pathomechanismen von Affektiven Störungen, G88412094
OpenAlex ID: W7115349300