CoolFace
Datasetpublic

open-misconceptions/miscon-data

Open Misconceptions A public catalogue of misconceptions with stable IDs. Each row is one record: a belief a learner could hold, its kind, the evidence pattern that reveals it (with a concrete example), discriminators against slips and neighbouring misconceptions, alignments to external schemes, and provenance. This dataset mirrors dist/miscon.jsonl from the tagged release of https://github.com/open-misconceptions/miscon-data. The canonical form of a record is its stable URI… See the full description on the dataset page: https://huggingface.co/datasets/open-misconceptions/miscon-data.

sourceHugging Facecc-by-4.0updated 15d agoView on Hugging Face
0likes77downloads
Dataset Card

Open Misconceptions

A public catalogue of misconceptions with stable IDs. Each row is one record: a belief a learner could hold, its kind, the evidence pattern that reveals it (with a concrete example), discriminators against slips and neighbouring misconceptions, alignments to external schemes, and provenance.

This dataset mirrors dist/miscon.jsonl from the tagged release of <https://github.com/open-misconceptions/miscon-data>. The canonical form of a record is its stable URI, for example <https://open-misconceptions.github.io/miscon-data/m/math.fractions.add-across>.

Provenance caveat

Most records are status: draft with trust: low: drafted from the cited literature and not yet reviewed. Check status, trust and reviews before treating a record as settled. reviewed records carry a human accept or an attested review; trust is computed from the reviews against the repository's reviewer registry.

Fields

FieldMeaning
idStable ID (math.fractions.add-across). Cite as miscon:<id>.
uriStable URI.
uuidOpaque identifier; the CASE CFItem.identifier.
version, status, trustRecord semver; lifecycle (draft, llm-reviewed, reviewed, deprecated, merged); computed trust (low, medium, high).
title, statement, notesShort label, the belief as the learner holds it, and optional notes such as likely origins.
kindMechanism: overgeneralization, undergeneralization, procedural-bug, notation-confusion, misapplied-analogy.
domain, about, level_band, localeSubject, concepts, education levels, language.
evidence_patternsList of {item_shape, signature, example{item, expected, response}}.
discriminatorsvs_slip and vs{<neighbour-id>: text}.
relationsconflicts_with, resolved_by, and provisional confusable_with, specializes, co_occurs_with, blocked_by.
alignments{scheme, code, uri, relation, note} into CCSS, progmiscon, etc.
provenance{sources[], origin, notes}.
reviews, history, licenseReviews (kind, by, date, scope, verdict), merge/supersede history and changelog, always CC-BY-4.0.

Full field documentation: <https://github.com/open-misconceptions/miscon-data/blob/main/schema/README.md>.

Load

python
from datasets import load_dataset
ds = load_dataset("open-misconceptions/miscon-data")

Cite

Concept DOI (latest release): https://doi.org/10.5281/zenodo.22416011. Each release also has a version DOI (v0.1.1: https://doi.org/10.5281/zenodo.22416012). Author and title metadata are in CITATION.cff in the repository.

Licence

CC BY 4.0. Attribute "Open Misconceptions (Vikram Maram and contributors)" and link to the repository or the record URI.