Compiled by: research agent, 2026-04-29 Purpose: Map what's empirically known about transfer of CT and digital skills training in adult/workforce populations (not K-12). Used to ground pedagogy.md and to scope the gap our project would fill.
Headline verdict
The empirical evidence base on adult computational thinking and digital-skills transfer is thin — far thinner than the K-12 literature. The asymmetry is stark and the doc should state it plainly.
What exists:
- One promising pilot outcome study (RAND 2024).
- One critical negative finding (Urban Institute: smartphone fluency does not transfer to office software).
- Large meta-analyses on K-12 CT transfer (Ye 2022, n = 55 studies; CT-STEM 2024, n = 37 studies, n = 7,832 students) with negligible adult representation.
- Suggestive evidence training helps (OECD PIAAC, program evaluations).
- Zero RCTs on adult CT transfer.
- Zero empirical studies on adult mental-model construction for digital systems.
Why the gap: research funding has concentrated on K-12 and higher education; the "digital natives" framing of the 2000s–2010s led researchers to underweight adult reskilling; adult basic education is chronically under-funded for outcome research.
Strongest empirical findings
A. Urban Institute 2019 — smartphone-to-office transfer failure
- Already in our research at
grey-literature/urban-institute-2019-foundational-digital-skills.md. - Provider-interview based (not a primary empirical study), but the finding circulates as the canonical example of adult digital-skill transfer failure.
- Relevance: anchors our mental-model design constraint. Smartphone fluency does not automatically extend to file systems, hierarchies, and persistence-based desktop work.
B. RAND 2024 — Computer Foundations for low-tech adults (pilot)
- Adults with limited prior computer experience.
- Daily computer use doubled; perceived proficiency rose; employment rates nearly tripled.
- Caveat: described as a "pilot"; sample size and control-group details not transparent in the abstract. Upper bound of available adult outcome evidence.
- URL: https://www.rand.org/pubs/research_reports/RRA3912-1.html
C. Ye et al. 2022 — meta-analysis of 55 CT-transfer studies
- K-12 and some higher-ed; adults negligible.
- Significant overall transfer effect; moderate near and far transfer.
- Reviewers explicitly note adults "received relatively little attention."
- URL: https://onlinelibrary.wiley.com/doi/abs/10.1111/jcal.12723
D. CT-STEM 2024 meta-analysis — 37 studies, n = 7,832
- Overwhelmingly K-12.
- Moderate overall transfer effect, both near and far.
- Stronger effects on cognitive than non-cognitive outcomes.
- URL: https://stemeducationjournal.springeropen.com/articles/10.1186/s40594-024-00498-z
E. Older-adults observational training — n = 59 (MDPI 2020)
- Behavior-modeling training improved attitude and adoption intent in older adults.
- Did not measure skill transfer to actual tasks. Affective/perceptual outcomes only.
- URL: https://www.mdpi.com/2071-1050/12/11/4555
F. Digital-divide systematic review (ERIC)
- Key finding: transfer to everyday life is dilutied by structural barriers (no home internet, unsuitable devices, cost), not just pedagogy.
- URL: https://files.eric.ed.gov/fulltext/EJ1363193.pdf
G. ProLiteracy — adult basic ed research review
- Identifies success factors and barriers in ABE digital integration. Does not report direct transfer outcomes.
- URL: https://www.proliteracy.org/resources/digital-literacy-and-technology-integration-in-adult-basic-skills-education-a-review-of-the-research/
H. OECD PIAAC — descriptive, not experimental
- Strong positive relationship between adult education participation and skills proficiency.
- Skills decline over the lifespan if not continuously practiced — implies that transfer to real life requires sustained use, not a one-shot training.
- Already cited in our pitch's Problem section.
Mental-model acquisition for digital systems — what we know
Even sparser than the transfer literature.
Chi 2008 — Three Types of Conceptual Change
- Three pathways: belief revision, mental-model transformation, categorical (ontological) shift.
- Highly relevant for our mental-model design constraint: file systems, persistence, and hierarchical organization may require categorical shifts for users who have only experienced phone-based, app-encapsulated state.
- Caveat: developed with school-age learners; no empirical studies test the framework with adult digital learners.
- URL: https://education.asu.edu/lcl/publications/chi-m-t-h-2008-three-types-conceptual-change-belief-revision-mental-model
NN/G — cloud-storage mental models
- Smartphone-first users map cloud services onto email-attachment mental models, lacking hierarchical/persistence understanding.
- User-testing synthesis, not a controlled study, but converges with Urban Institute.
- URL: https://www.nngroup.com/articles/cloud-storage/
Scaffolding framework (Belland)
- Six scaffolding strategies (cognitive + motivational): recruitment, reduction of degrees of freedom, direction maintenance, marking critical features, frustration control, demonstration.
- No adult-specific comparison studies found.
- URL: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3827669/
Counter-evidence and complications
- Skills decline without practice (OECD PIAAC). Training transfer is not just a pedagogical problem — it requires sustained opportunity for practice post-training.
- Affect and self-efficacy may matter more for adults than K-12. Multiple adult studies show motivational/affective change as the proximal outcome. Implication: our co-pilot must build confidence, not just skill.
- Structural barriers dwarf pedagogical ones. Home internet, device, cost. Pedagogy can't fix this; product design must account for it (offline modes? library distribution?).
- The "third-level digital divide" is real. Even with equal access, outcomes diverge based on prior knowledge, self-efficacy, and opportunity.
Strength-of-evidence by claim (this is the table to put in pedagogy.md)
| Claim | Evidence strength |
|---|---|
| Adults can develop digital skills through training | Strong (RAND, OECD, program evaluations) |
| Those skills transfer to novel contexts (far transfer) | Weak (no direct evidence; inferred from RAND employment outcomes; confounded) |
| Adults develop new mental models for digital systems | Extremely weak (theoretical frameworks only; no empirical studies) |
| Adult transfer is comparable to K-12 | Unknown (no direct comparisons) |
| CT training improves adult problem-solving | Not studied empirically in adults |
What this means for pedagogy.md and the pitch
This is the most important framing decision: be honest about the evidence base and turn the gap into a contribution claim. A pitch that pretends K-12 CT findings transfer to adults will get caught. A pitch that explicitly states "the adult digital-fluency transfer literature is empirically thin and our deployment is structured to generate evidence" is rigorous and credible.
The combination of:
- Strong K-12 CT-transfer meta-analyses (Ye 2022, CT-STEM 2024) — establishes the mechanism is real
- Robust pre-LLM ITS evidence on self-explanation transfer (d ≈ 0.33–0.55) — establishes our design moves work in some AI-tutored contexts
- Solid Bastani 2026 RCT in high-school Python — establishes adaptive AI tutoring works in a near-adjacent population
- The Urban Institute smartphone-to-office finding — establishes the specific failure mode we address
- The empirical gap on adult digital-fluency far transfer — establishes the contribution opportunity
...is enough to write a serious pedagogy doc. The honesty about the gap is the credibility move.