.png){kind=icon}
.png){kind=icon}
Return
Modeling meta-reasoning processes using diffusion and quantum random walk models
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jmp.2025.102952.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Meta-reasoning studies investigate the role of metacognitive processes in monitoring the success likelihood of an ongoing Reasoning task expected to require Longer Deliberation Time (RLDT), and accordingly, control further cognitive resource allocation to maximize success likelihood. A Meta-reasoning study may require participants to report their intermediate confidence judgment repeatedly within RLDT, e.g., a response that I am 70% confident that the problem is solvable, requested every 15 s. Based on existing Meta-reasoning studies, the current study first identified a set of observable Meta-reasoning phenomena on how intermediate confidence judgment evolves within RLDT and its impact on response choice and response time. Then, based on identified Meta-reasoning phenomena, certain computational features, serving as guidelines, were proposed to facilitate the construction and evaluation of random walk models describing these phenomena. The Markov Random-Walk formulation of the Drift-Diffusion Model (MR-DDM) and the Quantum Random-Walk Model (QRM) have been widely utilized to model response choice, response time, and intermediate and final confidence judgments in decision-making studies. Hence, the proposed computational features were utilized to evaluate the effectiveness of existing MR-DDM and QRM in describing meta-reasoning processes. Also, potential extensions of MR-DDM and QRM were identified for further empirical investigations. The current study also briefly reviewed an existing Item Response Theory (IRT) based extension of a continuous state continuous time Drift-Diffusion Model, named Q-Diffusion, that has been utilized to model RLDT without explicitly constraining the model to describe Meta-reasoning phenomena. Utilizing insights from Q-Diffusion and proposed computational features, the current study identified potential extensions of the MR-DDM for further empirical investigations.
Keywords:
Meta-reasoning
Confidence judgments
Response time
Drift-diffusion model
Quantum random-walk model
Journal
J
IF:
1.5
Papers:
24
Citations:
3.6K