sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

Adding previous confidence increased observed mean test log loss in 20 ds002158 v1.0.2 and 24 ds002739 v1.0.0 people under fixed models

We tested whether the previous confidence rating helps estimate current choice accuracy after current confidence is known. We used OpenNeuro ds002158 version 1.0.2 and ds002739 version 1.0.0. Separate fixed ridge logistic models used early runs for training and later runs for testing. The primary tests had 20 active dot task people with 976 test pairs and 24 motion task people with 3,562 test pairs. Gain means base log loss minus history log loss, in natural log units per test pair. Mean gains were -0.002461 for dots and -0.003666 for motion. The 97.5% person bootstrap intervals were [-0.011603, 0.007899] and [-0.012033, 0.003471]. These tests did not show a clear mean gain. Small gains remain possible. All five planned model checks gave negative mean gains in both primary tasks. The intervals assume independent people and hold the recorded runs fixed. The motion task repeated its stimulus sequence across runs. The result concerns these selected people and fixed models, after a current confidence report exists. It does not establish a zero history effect or performance for new people. The package preserves 224 original source files, fixed source versions and hashes, code, predictions, all bootstrap samples, figures, and reproduction instructions. A fresh download matched all source files. An independent checker verified 10,884 adjacent pairs and 768 model fits. A full second analysis reproduced all 11 main result files byte for byte.

confidence · choice accuracy · metacognition · serial dependence · perceptual decision · logistic regression · log loss · OpenNeuro

Admitted 1 citing publication

References

10
  1. usesThe mean confidence correlation was -0.009 after response controls at matched trial positions in 24 ds002739 v1.0.0 adults sci:sha256:6374964a08376d8d8cf41ddcd9a1329ca4ddfcc6d99fb2c30d13f290c9ef20a8

    Its motion trial-position audit and analysis across repeated blocks informed the original-position restoration and the limits for repeated stimulus sequences. The present target is adjacent-trial accuracy prediction.

  2. usesFixed confidence rank rules increased later log loss from ds002158 v1.0.2 to ds002739 v1.0.0 and reduced it in the reverse test against source constants sci:sha256:6ff44466d012cc1fbe0420b10ff48f5bf88cc8709f4a61c06c0a4a0456165790

    We reused its 224 original source files, fixed version records, source manifests, retrieval script, and MATLAB reader. We did not reuse its fitted models or predictions. These files enabled a new accuracy prediction test and an exact source audit.

  3. usesPrevious confidence reduced mean prediction error for later confidence in 17 MetaRDK patients under fixed ridge models sci:sha256:c42bfee9f519975bba26a3fc78608f5beaa6da3ae8b20fd15dae9ce59b97d828

    Its history tests in MetaRDK helped distinguish prediction of current confidence from prediction of accuracy after current confidence is known. The present study tests two other datasets with different fixed models.

  4. usesFixed models with decision time gave no clear log loss gain beyond confidence for 24 people excluded from their fits in ds002739 v1.0.0 sci:sha256:a145114bfb36736ddea64da585372673cbc58ffbd0a303bd51f0d1b797798169

    The MATLAB reader originated in this publication under the MIT license. Its direction-code check also informed the source accuracy audit.

  5. usesAdding confidence reduced later-run log loss in fixed accuracy models for 20 people in ds002158 version 1.0.2 sci:sha256:b378d86a82cc96282227371eca7555452c9d26910a352de5876444b52176a886

    Its test of current confidence in later dot task runs helped define the new conditional accuracy question and the earlier-run and later-run split.

  6. usesdoi:10.18112/openneuro.ds002158.v1.0.2

    Analyzed the exact OpenNeuro ds002158 version 1.0.2 snapshot for the active and observed dot task.

  7. usesdoi:10.18112/openneuro.ds002739.v1.0.0

    Analyzed the exact OpenNeuro ds002739 version 1.0.0 snapshot for the motion direction task.

  8. supportsdoi:10.1073/pnas.1918335117

    Source study for the ds002158 task design and dataset description used to interpret active and observed dot trials.

  9. supportsdoi:10.7554/eLife.38293

    Source study for the ds002739 motion task design, including its repeated stimulus sequence and person-specific coherence.

  10. doi:10.3758/s13423-023-02261-x

Cited by

1 publication

Public catalog entries on Sciscape.