Bibliographic record
Abstract
Reinforcement learning illustrates how people represent and learn large-scale environments for spatial navigation. However, it is unclear how “reachable” spaces are represented when performing manual tasks like chopping vegetables. To analyze the mechanisms underlying learning reachable space, we used inspiration from a recent study (de Cothi et al., 2022) to develop a haptic maze task where human participants reached to a target while avoiding invisible haptic obstacles in the environment. Participants reached with a robotic handle, generating contact forces to simulate the maze boundary, obstacle walls, and the floor, which supported the hand. Participants were blind to their hand position, while the target and maze boundary remained visible. Two conditions were implemented for each experiment variation; maze obstacles were invisible in the first condition, then became visible in the second. Audio feedback was given upon finding the target. We tested participants in 25 unique mazes, performing 10 trials within each maze. Two experimental procedures were employed: one with a fixed target and varying starting locations (18 participants), while the other had a fixed starting location and variable targets (10 participants). We simulated and compared the likelihoods of 3 different reinforcement learning models: model-based (MB), model-free (MF) and successor representation (SR). MB-simulated agents learned a map of the maze, planning the shortest route to the target. MF-simulated agents cached and updated the expected total future reward for each action using experience, pursuing the highest-reward actions. SR-simulated agents generated a “cognitive” (predictive) map between grids (states), integrating it with the target to plan actions. Results suggested that humans use a combination of MB and MF to learn reachable spaces. MB agents had higher likelihoods (better represented human behaviour) than MF agents for earlier trials, but MF had greater likelihoods for later trials. Meanwhile, SR learning was significantly worse at predicting participant behaviour.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".