Online library learning in human visual puzzle solving
arXiv:2603.23244v1 Announce Type: new Abstract: When learning a novel complex task, people often form efficient reusable abstractions that simplify future work, despite uncertainty about the future. We study this process in a visual puzzle task where participants define and reuse helpers -- intermediate constructions that capture repeating structure. In an online experiment, participants solved puzzles of increasing difficulty. Early on, they created many helpers, favouring completeness over efficiency. With experience, helper use became more selective and efficient, reflecting sensitivity to reuse and cost. Access to helpers enabled participants to solve puzzles that were otherwise difficult or impossible. Computational modelling shows that human decision times and number of operations used to complete a puzzle increase with search space estimated by a program induction model with library learning. In contrast, raw program length predicts failure but not effort. Together, these resul
arXiv:2603.23244v1 Announce Type: new Abstract: When learning a novel complex task, people often form efficient reusable abstractions that simplify future work, despite uncertainty about the future. We study this process in a visual puzzle task where participants define and reuse helpers -- intermediate constructions that capture repeating structure. In an online experiment, participants solved puzzles of increasing difficulty. Early on, they created many helpers, favouring completeness over efficiency. With experience, helper use became more selective and efficient, reflecting sensitivity to reuse and cost. Access to helpers enabled participants to solve puzzles that were otherwise difficult or impossible. Computational modelling shows that human decision times and number of operations used to complete a puzzle increase with search space estimated by a program induction model with library learning. In contrast, raw program length predicts failure but not effort. Together, these results point to online library learning as a core mechanism in human problem solving, allowing people to flexibly build, refine, and reuse abstractions as task demands grow.
Executive Summary
The article presents a compelling empirical study on online library learning as a cognitive mechanism in human visual puzzle solving. Participants, as they engaged with progressively complex puzzles, initially created numerous helpers to capture structural patterns, prioritizing completeness over efficiency. With repeated exposure, their helper usage became more selective, demonstrating adaptive efficiency and sensitivity to reuse costs. Computational modeling corroborated that human decision times and operational efforts correlate with a program induction model incorporating library learning, indicating that abstracted representations significantly influence cognitive effort. Importantly, the study reveals that access to reusable abstractions enables solution of otherwise intractable problems, positioning online library learning as a central pillar in adaptive problem-solving behavior. The findings bridge cognitive psychology and computational modeling, offering empirical validation for the role of abstraction in human cognition.
Key Points
- ▸ Initial helper creation reflects a preference for completeness over efficiency in novel tasks.
- ▸ With experience, helper usage evolves toward selective efficiency, indicating adaptive learning.
- ▸ Computational modeling validates that library learning impacts decision time and operational effort, validating cognitive abstraction as a measurable phenomenon.
Merits
Empirical Validation
The study provides robust experimental data supporting the concept of online library learning as a real cognitive mechanism, not merely theoretical.
Computational Correlation
Modeling of decision time and operations aligns with abstraction dynamics, offering interdisciplinary credibility.
Demerits
Scope Limitation
The study is confined to visual puzzles; generalizability to other domains (e.g., mathematical or linguistic puzzles) remains unexamined.
Model Constraint
The computational model, while informative, does not fully account for individual variability in cognitive processing speed or prior knowledge.
Expert Commentary
This paper makes a significant contribution to the understanding of human cognitive adaptation through the lens of reusable abstraction. The longitudinal shift from quantity-driven to quality-driven abstraction creation is particularly noteworthy, as it mirrors evolutionary principles of efficiency optimization under constraint. The authors effectively integrate empirical observation with computational simulation, achieving a rare synthesis between behavioral science and algorithmic modeling. Moreover, the implication that access to abstractions enables solutions to previously unsolvable problems suggests a potential paradigm shift in cognitive intervention design—potentially applicable to mental health, education, or even legal reasoning where complex pattern recognition is critical. While the study’s domain specificity warrants caution, its methodological rigor and conceptual clarity elevate it to a benchmark for future research in human cognition. The next step would be to extend this framework to cross-domain tasks, particularly in domains requiring latent structure identification, to assess the universality of library learning as a mechanism.
Recommendations
- ✓ Future research should replicate the study with cross-domain puzzle types (e.g., mathematical, linguistic, spatial) to test the generality of library learning.
- ✓ Develop educational interventions based on the findings to assess impact on student performance in complex problem-solving tasks.
Sources
Original: arXiv - cs.AI
