Skip to content
Open access

Working memory adaptively recomputes plans amid distraction

Aug 2026 · bioRxiv · 0 citations
Biology

Abstract

Navigating daily tasks requires working memory to retain information, formulate plans, and execute goal-directed actions. While frequent distractions may momentarily disrupt planned actions, individuals typically exhibit the cognitive resilience necessary to regroup and resume goal-directed behavior. How people adapt to these perturbations in dynamic environments remains unclear. To address this, we developed a working memory paradigm inspired by the arcade game Snake, in which participants navigated an agent to collect memorized targets while EEG and eye-tracking data were recorded. On each trial, participants encoded the location of one or two targets (apples), maintained them during a delay, and then guided a virtual agent (snake) to collect them. Critically, on 50% of trials, a secondary task introduced high-value novel targets that required immediate pursuit; these interruptions altered the agent’s spatial position, necessitating a re-evaluation of the original plan upon returning to the primary goal. Behaviorally, participants prioritized the target proximal to the agent and flexibly revised their collection order after distraction according to the agent’s updated position. Eye-movement analyses showed structured gaze sweeps during encoding that were consistent with efficient route planning. Representational similarity analyses of EEG data revealed that agent-centered object vectors were represented more strongly than absolute object locations during the delay. Following distraction, object vectors were recomputed relative to the agent’s new position, and representational strength was redistributed when the action plan was updated. Whereas neural representations were biased toward the distal, future-relevant target, gaze was preferentially directed toward the proximal, immediately relevant target, suggesting complementary roles for internal attention and external attention in multistep planning. During distraction, backward gaze sweeps revisited memorized target locations, indicating that replanning unfolded dynamically while the intervening task was still ongoing. These data demonstrate that prospective neural coding and attentional sampling coordinate to adapt behaviors to the shifting demands of goal pursuit in dynamic environments.

Read PDF

Similar papers

Oct 2026

Temporal expectation triggers priority shifts in working memory that lead to forgetting.

Working memory (WM) flexibly prioritizes task-relevant information over short intervals, serving as a critical gateway to long-term memory (LTM). However, how this dynamic prioritization influences what is ultimately retained in LTM remains unclear. Here, we investigated the long-term consequences of internally guided...

Zi-Yi Duan, Zi-Yao Zhang, J. Lewis-Peacock · 0 citations
Preprint Oct 2026

ProactiveVLA: Augmenting Embodied Memory through Proactive Environment Exploration

Rapid adaptation to a new environment requires a robot to acquire useful knowledge about local objects, states, and interactions from limited experience. Systems that combine a reasoning agent with a frozen vision-language-action model (VLA) can adapt through execution feedback and memory, making the choice of experien...

Shi-Zuo Tian, Hao-Dong Luo, Yu-Tong Li et al. · 0 citations
Open access Sep 2026

Spatial goals trap hippocampal replay

Hippocampal replay is proposed to support memory and planning by preferentially reactivating behaviorally significant locations1,2. Although learning can alter which locations are reactivated and how frequently 3–7, whether it also alters the dynamics by which replay propagates through the cognitive map remains unknown...

Caitlin S. Mallory, John Widloski, David J. Foster · 0 citations
Open access 2026

Working Memory Network Task Reveals Brain Areas Involved in Goal-Directed Retention

Working Memory (WM) provides temporary storage and processing of information to enable goal-directed behavior. According to the cognits framework, WM arises from the interaction of perceptual and executive neuronal ensembles, with the latest contributing to the maintenance and prioritization of goal-relevant informatio...

S. Mening, T. Fedele, N. Otstavnov · 0 citations
#artificial intelligence Preprint Sep 2026

When Successful Memories Mislead Embodied Agents:Memory Adaption For Task-Conditioned Execution

Memory Adaptation for Task-Conditioned Execution (MATE) is introduced, a deterministic post-retrieval procedure that converts trajectories into execution-oriented memory and supports memory adaptation as a distinct stage between retrieval and embodied execution.

Quan-Quan Li, Hong-Bo Zhang, Yi-He Chi et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.