Agency
Did the learner set goals, monitor uncertainty, justify choices, and revise with evidence—not merely report feeling in control?
Systematic literature map · through August 13, 2026
THE FULL
COGNITIVE OFFLOADING
REPOSITORY.
Every paper in the source corpus, made easier to search, compare, question, and critique—without hiding access limits or review status.
RESEARCH TERRAIN262 TOTAL
01Education147
02Work & expertise70
03Cognition & society35
04Health & medicine10
TINA’S READING RULE
Preserve the useful finding. Name where the evidence ends. Measure what should come next.
262strict inclusions
104full text inspected
158abstract only
99preprint / unreviewed
SOURCE COVERAGE
3,311 raw records → 2,192 deduplicated → 262 strict inclusions
Evidence library
SEARCH WIDELY.
READ SKEPTICALLY.
The original corpus had breadth. This version adds comparison, progressive disclosure, and an honest appraisal layer to every record.
Showing 24 of 262 papersOpen any paper for critique + source
Source-grounded evaluation
WHAT WOULD
TINA’S LENS
REQUIRE?
Margin now evaluates every question against Tina R. Austin’s The Learner Is Not a Constant and her human–AI evaluation field guide. It does not imitate Tina. It makes the published standard visible.
Durable capability
What remained when support changed, failed, or disappeared: retention, transfer, calibration, recovery, or independent performance?
Human behavior
What did the person verify, accept, reject, revise, retrieve, explain, or delegate in response to the model?
Trajectory
Did the design follow changing behavior, context, treatment, and capability across enough observations to test its mechanism?
Ground truth
What independent learner, educator, assessor, institutional, or community evidence widened the model’s narrow view?
Method & limits
BREADTH
WITHOUT BLUFFING.
The source searched arXiv, OpenAlex, Crossref, Semantic Scholar, and PubMed with exact-term and close-variant query families. Inclusion required AI plus cognitive offloading, cognitive outsourcing, deskilling, over-reliance, or metacognitive laziness to be central—not merely cited in passing.
“Full text” means an openly accessible paper was extracted and inspected. “Abstract only” means the summary is bounded to accessible metadata. Margin’s Tina-aligned verdicts are rule-based syntheses of repository fields and the supplied framework—not live responses from Tina and not completed formal risk-of-bias reviews.
33 API requests/query pagesFive-pillar Tina lensUpdated Aug 2026