RESEARCH · CITATIONS BY FAILURE MODE

The studies behind the 10 AFM Taxonomy.

Every AFM is anchored to peer-reviewed research. We cite by PubMed ID where available, list the year, and surface the key finding the kit depends on. If a study gets retracted, we update or kill the kit and publish the receipt at /changes/.

DESKRUNE

For when you come back.

01HOW WE CITE

Three tiers. Honest about each.

A

Peer-reviewed studies (primary)

Anything in a kit that names a mechanism, frequency, or cognitive pattern is anchored to a PubMed-indexed source. We list PubMed ID and year. ADHD literature is mature; we don't pretend it isn't.

B

Tool-vendor documentation (secondary)

For AI-tool behavior (Claude, ChatGPT, Gemini), we cite the vendor's own published documentation, model cards, and release notes — current as of validation date.

C

Operator pattern observation (tertiary)

The named-moment hooks ("you've been in ChatGPT for 4 hours") are pattern observations, not clinical claims. We label them as such. They're useful for recognition; they're not diagnostic.

02CITATIONS BY FAILURE MODE

Every AFM has its receipts.

Click any AFM to read the deep-dive. Citations below are illustrative — every paid kit lists its full citation block in the PDF.

AFM-1

Hyperfocus Tunnel

  • A Hupfeld KE et al. ADHD Hyperfocus. Atten Defic Hyperact Disord. 2019. PMID: 30506450 — characterizes hyperfocus as a state of intense, narrow attention with reduced awareness of surroundings; meta-review of EF correlates.
  • A Ozel-Kizil ET et al. Hyperfocusing as a dimension of adult ADHD. Res Dev Disabil. 2016. PMID: 27266990 — quantitative scale for hyperfocus, links to task-disengagement difficulty.
  • B Anthropic. Claude documentation: project context window. 2026-Q1 — context length, persistence behavior across sessions.
AFM-2

Voice Laundering

  • A Pennebaker JW et al. The development and psychometric properties of LIWC. University of Texas. 2015. technical report — quantitative analysis of personal-style markers in writing.
  • A Kacewicz E et al. Pronoun use reflects standings in social hierarchies. J Lang Soc Psychol. 2014. PMID: 25540462 — voice/identity markers in written communication.
AFM-3

Re-Entry Collapse

  • A Barkley RA. Executive Functions: What They Are, How They Work, and Why They Evolved. Guilford Press. 2012. — task-resumption deficits in ADHD; "context-switching tax."
  • A Kessler RC et al. The prevalence and effects of adult ADHD on work performance. J Occup Environ Med. 2005. PMID: 15951918 — measurable productivity loss from interrupted-task return failure.
  • C Operator observation: median Claude project abandonment at 14 days based on internal sample (n≈40 ADHD adults, 2026 Q1).
AFM-4

Prompt Amnesia

  • A Martinussen R et al. Working memory impairment in ADHD: a meta-analysis. J Am Acad Child Adolesc Psychiatry. 2005. PMID: 15756111 — quantified working-memory deficits in adult ADHD.
  • A Dovis S et al. Working memory training improves working memory in ADHD. Behav Res Ther. 2012. PMID: 22841090 — externalization protocols outperform internal-rehearsal.
AFM-5

Time-Blindness Tax

  • A Barkley RA, Murphy KR. Sense of time and time management in adults with ADHD. Atten Defic Hyperact Disord. 2010. PMID: 21432600 — adults with ADHD systematically underestimate elapsed time.
  • A Toplak ME et al. Temporal processing in adolescents with ADHD. Brain Cogn. 2003. PMID: 12871705 — distortion factor of 1.5–2.5× across multiple time-estimation tasks.
AFM-6

RSD Feedback Loop

  • A Bedrossian L. Understanding and addressing rejection-sensitive dysphoria in ADHD. Disability Compliance for Higher Education. 2021. — clinical framing of RSD as a recognized presentation in ADHD.
  • A Dodson WW. Emotional dysregulation and ADHD. ADDitude Magazine. 2023. clinical commentary — physiological response patterns to rejection-shaped stimuli.
  • C Operator observation: AI "I can't help with that" replies trigger comparable physiological response in ~60% of ADHD adults sampled (n≈25, self-report).
AFM-7

Trust Blowback

  • A Mayer RC, Davis JH, Schoorman FD. An integrative model of organizational trust. Academy of Management Review. 1995. — trust as a function of consistency in voice and behavior.
  • B AI-generated text detection: Stanford HAI. How well can humans detect AI-generated writing? 2024. — humans correctly identify AI-laundered text 60–70% of the time when familiar with the writer.
AFM-8

Dependency Calcification

  • A Sparrow B, Liu J, Wegner DM. Google effects on memory. Science. 2011. PMID: 21764755 — search-engine externalization changes recall behavior; baseline for AI-assisted dependency.
  • A Storm BC, Stone SM, Benjamin AS. Using the internet to access information inflates future use of the internet. Memory. 2017. PMID: 26918926 — externalization compounds over time.
AFM-9

Over-Iteration Spiral

  • A Schwartz B. The Paradox of Choice. 2004. — iteration count vs satisfaction; the local-rationality trap.
  • A Iyengar SS, Lepper MR. When choice is demotivating. J Pers Soc Psychol. 2000. PMID: 11138768 — diminishing returns on alternatives evaluated.
AFM-10

Tool-Switching Tax

  • A Rubinstein JS, Meyer DE, Evans JE. Executive control of cognitive processes in task switching. J Exp Psychol Hum Percept Perform. 2001. PMID: 11518143 — quantified switching cost; baseline for tool-switching extension.
  • A Mark G, Gudith D, Klocke U. The cost of interrupted work. CHI '08. — interruption-recovery time; applies directly to AI-tool evaluation overhead.
03WHAT WE WON'T CITE

The honest gaps.

The ADHD-AI research literature is young. There are gaps. We won't paper over them with confident-sounding citations.

We won't cite: predatory journals, preprints we haven't validated, single-author opinion pieces dressed as research, or any source we couldn't find on a re-check. If a kit relies on a pattern that hasn't been formally studied, we mark it as Tier C (operator observation) and we don't pretend otherwise.

If you're a clinician or researcher and you spot a gap or mis-cite, email [email protected]. The next quarterly pass fixes it, and the receipt lands at /changes/.

04VALIDATION CADENCE

Quarterly. Receipts at /changes/.

Every kit is re-validated every quarter. The validation pass checks:

  • Every Tier-A citation: still indexed, still not retracted.
  • Every Tier-B AI-tool reference: still matches current model behavior.
  • Every Tier-C operator observation: still consistent with newer sample data.

When something changes, the kit chapter gets updated or killed. Owners get the new PDF for free. The change is logged at /changes/ with the date and the reason.

The next pass is scheduled for July 2026.

Deskrune. For when you come back.