Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Differences between Neurodivergent and Neurotypical Software Engineers: Analyzing the 2022 Stack Overflow Survey

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Comparing 2022 Stack Overflow survey responses, neurodivergent software engineers show few significant differences from neurotypical peers, with the clearest effects appearing for ADHD engineers in hybrid work.

desk verdict Useful, honest baseline for ND-NT comparisons in SE, but the abstract oversells and the neurotypical group's exclusion criteria need clarification. read the letter →

arxiv 2506.03840 v1 pith:NDRVSISN submitted 2025-06-04 cs.SE

Pragya Verma , Marcos Vinicius Cruz , Grischa Liebel This is my paper
classification cs.SE
keywords neurodiversitysoftwareengineeringStackOverflowsurveyADHDautismspectrumdisorderdyslexiaknowledgesharingworkplaceinclusion
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish whether employed neurodivergent software engineers—those with autism, ADHD, or dyslexia—experience knowledge sharing, interaction, and information finding at work differently from neurotypical colleagues. Using 2022 Stack Overflow survey data, the authors compare 374 ASD, 1305 ADHD, and 363 dyslexic professionals to equal-size neurotypical samples, both random and matched by work mode and experience. They find very few statistically significant differences, and all have small effect sizes. The clearest pattern is that engineers with ADHD, especially in hybrid work, report more workflow disruptions from waiting on answers and fewer interactions beyond their team. The paper offers this as a conservative baseline: if real differences exist, the survey's design likely underestimates them.

What carries the argument

The analysis rests on ten survey items (K1–K7 on knowledge and information seeking; F1–F3 on interaction frequency), the single neurodiversity self-report filter that isolates single-condition groups, and Mann-Whitney U tests with Bonferroni-Holm correction. The paired sampling by work mode and experience is the mechanism that isolates the conditions under which differences appear.

What would settle it

Re-run the comparison using a clinically validated neurodiversity screening or diagnostic confirmation on the same sample; if the resulting effect sizes grow beyond small, or the ADHD K7/F2 differences appear in remote and in-person groups too, the paper's null-leaning conclusion is overturned.

Watch

Extended reading notes

Core claim

The paper's central claim is that employed neurodivergent software engineers do not differ substantively from neurotypical peers on the ten knowledge-sharing, interaction, and information-finding questions in the 2022 Stack Overflow survey. After correcting for multiple testing, only a handful of comparisons reach significance, and effect sizes are small (r between -0.07 and -0.22). The most consistent finding is for ADHD: these engineers report that waiting on answers disrupts their workflow (K7) and that they interact less with people outside their immediate team (F2), an effect that persists and sharpens in the hybrid-work subsample. For ASD, one unpaired comparison (K4) shows more difficulty finding answers with existing tools. The authors read these results as evidence that while ND professionals face real difficulties, they are not substantially different from NT professionals on these measures, and they caution that design features likely make the estimates conservative.

Load-bearing premise

The survey's one self-report question, after filtering out co-morbidities, correctly separates neurodivergent from neurotypical respondents; if many neurotypical respondents are undiagnosed neurodivergent, every comparison is biased toward no difference.

Editorial extensions

If this is right

  • If the small-effect finding holds, workplace context and accommodations likely matter more than a neurotype label for day-to-day knowledge work.
  • ADHD engineers in hybrid settings are the group most worth targeted support: reducing wait times for answers could cut workflow interruptions.
  • The absence of differences in fully remote and full in-person comparisons means work mode should be a controlled variable in future neurodiversity-in-SE research.
  • Future survey-based studies need validated diagnostic screening, not a single self-report item, to detect effects that this baseline could not.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • This editor infers that the study's restriction to employed engineers likely removes the most affected individuals; re-running with unemployed or underemployed ND developers could show larger effects than reported here.
  • The K7 finding suggests a concrete, testable mechanism: for ADHD engineers, the latency of asynchronous question-answer channels is disproportionately disruptive; a tool-log study measuring interruption frequency around wait times could confirm it.
  • The continent imbalance between ND and NT samples (ND respondents skewed European, ADHD respondents North American) may reflect diagnostic culture rather than workplace experience; matching by country could change which differences survive.
  • Because the survey's Likert items are coarse, real differences could be masked; a purpose-built instrument with finer scales and direct measures of accommodation use is the natural next test.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper analyzes the 2022 Stack Overflow Developer Survey to compare professional software engineers with self-reported ASD (n=374), ADHD (n=1305), and dyslexia (n=363) against neurotypical engineers on ten questions concerning knowledge sharing, interaction, and information finding. The authors use Mann-Whitney U tests with Bonferroni-Holm correction applied across all tests within each condition, and they compare the groups under three sampling schemes: unpaired random samples, samples matched on work mode, and samples matched on years of work experience. The main finding is that few statistically significant differences emerge and the significant ones have small effect sizes, with the most consistent results being that ADHD engineers, particularly in hybrid work, report more workflow interruptions when waiting for answers (K7) and fewer interactions with people outside their team (F2). The paper interprets these results as a conservative baseline for future research and includes a detailed validity-threats discussion.

Significance. If the results are valid, this is a valuable large-scale quantitative complement to the mainly qualitative neurodiversity literature in software engineering. The study is unusually transparent: the filtered data and analysis scripts are available on Zenodo, the source data are public, and the statistical pipeline is clearly described. The ADHD-related differences in K7 and F2 replicate across the unpaired and hybrid-matched samples with corrected p-values below 0.01, providing a concrete, checkable result. The paper also deserves credit for an honest validity-threats section. However, the scientific value hinges critically on the clean separation of the neurotypical comparison group, because the headline result is an absence of differences; the manuscript's ambiguity on this point is the main concern.

major comments (3)
  1. [III-A1, V] The construction of the neurotypical comparison group is ambiguous and potentially load-bearing for the paper's null-leaning conclusions. Section III-A1 defines the NT group as 'no neurodiversity (none of the specified conditions)', while Section V-C states that the authors filtered 'people with mental health issues such as anxiety or depression' out of the ND groups, but nowhere states that the same exclusion was applied to the NT group. Since the survey recorded 'neurodiversity and/or any emotional or anxiety disorder' in a single question, a response of 'none of the specified conditions' may still include respondents with anxiety or depression. Such respondents would plausibly report more workflow interruptions and fewer social interactions, diluting any ND-NT differences and biasing the analysis toward the observed null results. The authors should clarify the exact survey item and their filtering code, and if anxiety/depression respondents are retained in the NT group, re-run all comparisons with an NT group that excludes every listed condition. In addition, Section V acknowledges 'undiagnosed conditions in our NT sample' but does not quantify this threat; a sensitivity analysis that estimates the robustness of the findings to plausible contamination rates would materially strengthen the claim.
  2. [IV-C, IV-D] The null results in several strata are difficult to interpret because per-stratum sample sizes are not reported and some are very small. Section III-A1 lists, for the work-mode strata, in-person groups of n=34 (dyslexia), n=133 (ADHD), and n=41 (ASD); with these sizes and a Bonferroni-Holm correction across multiple tests per condition, the Mann-Whitney U test has limited power to detect small effects. The paper's aggregate conclusion that 'we find only few significant differences' relies on these underpowered comparisons as much as on the well-powered unpaired and hybrid analyses. The authors should report the sample size for every comparison in the paper, provide a power analysis or the minimum detectable effect size for each stratum, and temper the conclusions accordingly when power is inadequate.
  3. [III-B, Tables I-III] The effect-size measure used in Tables I-III is not defined. The tables report values in parentheses and label them as 'small' or 'medium', but the manuscript does not state whether these are rank-biserial correlations, Z/sqrt(N), or another statistic, nor the thresholds used to classify them. Because the claim that the significant differences are 'small' is central to the paper's interpretation, the authors must define the measure, give the classification thresholds, and ideally report confidence intervals for the effect sizes.
minor comments (5)
  1. [III-A1, IV] The term 'paired sampling' is misleading: the authors frequency-match on work mode and work experience but still apply unpaired Mann-Whitney U tests; consider using 'matched sampling' throughout.
  2. [Abstract, IV-D1] The abstract and Section I say the observed effects are 'small', but Section IV-D1 reports a 'medium effect size' (-0.219) for K5 in the 0-5 years dyslexia stratum; harmonize this wording.
  3. [III-A1, V] Section V says the study 'included only employed and retired developers', but Section III-A1 restricts the sample to full-time employed respondents; clarify which population was actually analyzed.
  4. [IV-B] In the description of the significant ASD K4 finding, the direction of the effect is not tied to its sign; state explicitly that the negative effect size corresponds to lower agreement with being able to quickly find answers.
  5. [Tables I-III] The tables report corrected p-values of '1.00' for many non-significant comparisons; consider reporting the uncorrected p-values in an appendix so that readers can gauge the raw evidence.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the analysis is a secondary quantitative study of the externally collected 2022 Stack Overflow survey, and the few author self-citations are background only.

full rationale

The paper's central claim—that there are few significant differences between neurodivergent and neurotypical engineers on ten knowledge/interaction items—is obtained by applying Mann-Whitney U tests with Bonferroni-Holm correction to answers taken from the publicly available 2022 Stack Overflow Developer Survey. The ND and NT groups are defined by the survey's own self-report question plus explicit filtering (Section III-A1), and the outcome questions (K1-K7, F1-F3) are fixed survey items. Nothing in the statistical comparison is fitted to the authors' own definitions or to a prior result of theirs; the results are computed directly from the external dataset. The only author-overlapping citations, [4] and [7] (Liebel and Gama), appear in the related-work discussion and as qualitative context for why accommodations and environment matter; they are not the evidentiary basis for the K7/F2 findings. The acknowledged limitation about undiagnosed or undisclosed neurodivergent respondents in the NT group (Section V) is a construct-validity threat regarding who is in each comparison group, not a circular step: it does not make the reported differences equal to the paper's inputs by construction. Accordingly, no specific circular reduction can be quoted, and the appropriate finding is no significant circularity.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The central claim rests entirely on external survey data; no free parameters are fitted to produce the result, and no entities are invented. The load-bearing premises are: (1) the survey's neurodiversity self-report plus single-condition filtering separates ND from NT groups, (2) questions K1-K7 and F1-F3 validly measure knowledge sharing, interaction, and information finding, and (3) respondents are independent, satisfying the Mann-Whitney U assumption. Premises (1) and (2) are explicitly flagged by the authors as validity threats; premise (3) is standard for this survey design.

assumptions (4)
  • domain assumption The 2022 Stack Overflow survey's neurodiversity self-report question, combined with the authors' filtering to single-condition groups, correctly classifies respondents as neurodivergent or neurotypical.
    Every significance test depends on this grouping. Section III-A1 describes the filtering; Section V acknowledges that undiagnosed or undisclosed conditions in the NT sample could affect results.
  • domain assumption Questions K1-K7 and F1-F3 validly measure knowledge sharing, interaction, and finding information at work.
    Section III-C3 raises construct validity as an explicit threat; the authors note ambiguous wording, for example whether 'interaction' means work-related or social interaction, and that they had no control over question design.
  • standard math Respondents are independent, satisfying the key assumption of the Mann-Whitney U test.
    Section III-B states that survey participants are independent because they are individual respondents, which is standard for this analysis.
  • domain assumption Pairing on work mode and years of work experience controls for the main confounding variables.
    Section III-A1 motivates the paired sampling; Sections IV-C and IV-D report per-stratum results. Other confounders such as accommodations and condition severity are unmeasured and acknowledged in Section V.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Differences between Neurodivergent and Neurotypical Software Engineers: Analyzing the 2022 Stack Overflow Survey." pith.science (2026). https://pith.science/paper/NDRVSISN

@misc{pith2026250603840,
  author       = {Pith},
  title        = {Pith review of: Differences between Neurodivergent and Neurotypical Software Engineers: Analyzing the 2022 Stack Overflow Survey},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NDRVSISN}},
  note         = {Machine review of arXiv:2506.03840}
}
read the original abstract

Neurodiversity describes variation in brain function among people, including common conditions such as Autism spectrum disorder (ASD), Attention deficit hyperactivity disorder (ADHD), and dyslexia. While Software Engineering (SE) literature has started to explore the experiences of neurodivergent software engineers, there is a lack of research that compares their challenges to those of neurotypical software engineers. To address this gap, we analyze existing data from the 2022 Stack Overflow Developer survey that collected data on neurodiversity. We quantitatively compare the answers of professional engineers with ASD (n=374), ADHD (n=1305), and dyslexia (n=363) with neurotypical engineers. Our findings indicate that neurodivergent engineers face more difficulties than neurotypical engineers. Specifically, engineers with ADHD report that they face more interruptions caused by waiting for answers, and that they less frequently interact with individuals outside their team. This study provides a baseline for future research comparing neurodivergent engineers with neurotypical ones. Several factors in the Stack Overflow survey and in our analysis are likely to lead to conservative estimates of the actual effects between neurodivergent and neurotypical engineers, e.g., the effects of the COVID-19 pandemic and our focus on employed professionals.

Figures

Figures reproduced from arXiv: 2506.03840 by the authors.

Figure 3
Figure 3. Daily time answering questions in the ASD sample and the corre [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Daily time searching in the Dyslexia sample and the corresponding [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 1
Figure 1. Daily time answering questions in the Dyslexia sample and the [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figures from the paper (3 more)
Figure 2
Figure 2. Figure 2: Daily time answering questions in the ADHD sample and the [PITH_FULL_IMAGE:figures/full_fig_p005_2.png]
Figure 5
Figure 5. Figure 5: Daily time searching in the ADHD sample and the corresponding [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Daily time searching in the ASD sample and the corresponding [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Tether: A Personalized Support Assistant for Software Engineers with ADHD

    cs.SE 2025-09 conditional novelty 6.0 of 10

    An LLM-powered desktop assistant combining activity monitoring, retrieval-augmented generation, and gamification to support software engineers with ADHD, validated only through the authors' self-use.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [1]

    Diamond in the rough? an “empty review

    N. Doyle and A. McDowall, “Diamond in the rough? an “empty review” of research into “neurodiversity” and a road map for developing the inclusion agenda,” Equality, Diversity and Inclusion: An International Journal, vol. 41, no. 3, pp. 352–382, 2021

    work page 2021

  2. [2]

    Odd people in: The birth of community amongst people on the “autistic spectrum

    J. Singer, “Odd people in: The birth of community amongst people on the “autistic spectrum”,” 1998

    work page 1998

  3. [3]

    Neurodiversity at work: a biopsychosocial model and the impact on working adults,

    N. Doyle, “Neurodiversity at work: a biopsychosocial model and the impact on working adults,” British medical bulletin, vol. 135, no. 1, pp. 108–125, 2020

    work page 2020

  4. [4]

    Challenges, strengths, and strategies of software engineers with adhd: A case study,

    G. Liebel, N. Langlois, and K. Gama, “Challenges, strengths, and strategies of software engineers with adhd: A case study,” in Proceedings of the 46th International Conference on Software Engineering: Software Engineering in Society , ser. ICSE-SEIS’24. New York, NY , USA: Association for Computing Machinery, 2024, p. 57–68. [Online]. Available: https://d...

    work page arXiv 2024

  5. [5]

    Inclusion of individuals with autism spectrum disorder in software engineering,

    G. M ´arquez, M. Pacheco, H. Astudillo, C. Taramasco, and E. Calvo, “Inclusion of individuals with autism spectrum disorder in software engineering,” Information and Software Technology , p. 107434, 2024

    2024

  6. [6]

    Eye tracking analysis of computer program comprehension in programmers with dyslexia,

    I. McChesney and R. Bond, “Eye tracking analysis of computer program comprehension in programmers with dyslexia,” Empirical Software Engineering, vol. 24, no. 3, pp. 1109–1154, 2019

    work page 2019

  7. [7]

    A socio- technical grounded theory on the effect of cognitive dysfunctions in the performance of software developers with adhd and autism,

    K. Gama, G. Liebel, M. Goul ˜ao, A. Lacerda, and C. Lacerda, “A socio- technical grounded theory on the effect of cognitive dysfunctions in the performance of software developers with adhd and autism,” in Pro- ceedings of the 47th International Conference on Software Engineering: Software Engineering in Society , ser. ICSE-SEIS’25. New York, NY , USA: Ass...

    work page 2025

  8. [8]

    Understanding the chal- lenges faced by neurodiverse software engineering employees: Towards a more inclusive and productive technical workforce,

    M. R. Morris, A. Begel, and B. Wiedermann, “Understanding the chal- lenges faced by neurodiverse software engineering employees: Towards a more inclusive and productive technical workforce,” in Proceedings of the 17th International ACM SIGACCESS Conference on computers & accessibility, 2015, pp. 173–184

    2015

Show all 26 references
  1. [9]

    Dataset: Differences between neurodivergent and neu- rotypical software engineers: Analyzing the 2022 stack overflow survey,

    Anonymous, “Dataset: Differences between neurodivergent and neu- rotypical software engineers: Analyzing the 2022 stack overflow survey,” Jan. 2025. [Online]. Available: https://doi.org/10.5281/zenodo.14779344

    2022 doi

  2. [10]

    Observations on the linear order of program code reading patterns in programmers with dyslexia,

    I. McChesney and R. Bond, “Observations on the linear order of program code reading patterns in programmers with dyslexia,” in Proceedings of the Evaluation and Assessment in Software Engineering , 2020, pp. 81– 89

    2020

  3. [11]

    Towards accessible remote work: Understanding work-from-home practices of neurodi- vergent professionals,

    M. Das, J. Tang, K. E. Ringland, and A. M. Piper, “Towards accessible remote work: Understanding work-from-home practices of neurodi- vergent professionals,” Proceedings of the ACM on Human-Computer Interaction, vol. 5, no. CSCW1, pp. 1–30, 2021

    2021

  4. [12]

    Managing stress: The needs of autistic adults in video calling,

    A. Zolyomi, A. Begel, J. F. Waldern, J. Tang, M. Barnett, E. Cutrell, D. McDuff, S. Andrist, and M. R. Morris, “Managing stress: The needs of autistic adults in video calling,” Proceedings of the ACM on Human- Computer Interaction, vol. 3, no. CSCW, pp. 1–29, 2019

    2019

  5. [13]

    Understand- ing how adhd affects visual information processing,

    Y . Alqahtani, M. McGuire, J. Chakraborty, and J. H. Feng, “Understand- ing how adhd affects visual information processing,” in International Conference on Human-Computer Interaction . Springer, 2019, pp. 23– 31

    2019

  6. [14]

    Adhd at the workplace: Adhd symptoms, diagnostic status, and work-related functioning,

    A. B. Fuermaier, L. Tucha, M. Butzbach, M. Weisbrod, S. Aschenbren- ner, and O. Tucha, “Adhd at the workplace: Adhd symptoms, diagnostic status, and work-related functioning,” Journal of Neural Transmission , vol. 128, pp. 1021–1031, 2021

    2021

  7. [15]

    Fatigue in an adult attention deficit hyperactivity disorder population: A trans- diagnostic approach,

    D. C. Rogers, A. J. Dittner, K. A. Rimes, and T. Chalder, “Fatigue in an adult attention deficit hyperactivity disorder population: A trans- diagnostic approach,” British Journal of Clinical Psychology , vol. 56, no. 1, pp. 33–52, 2017

    2017

  8. [16]

    Perceived stress and adhd symptoms in adults,

    M. A. Combs, W. H. Canu, J. J. Broman-Fulks, C. A. Rocheleau, and D. C. Nieman, “Perceived stress and adhd symptoms in adults,” Journal of Attention Disorders , vol. 19, no. 5, pp. 425–434, 2015

    2015

  9. [17]

    The relationships between workaholism and symptoms of psychiatric disorders: A large-scale cross-sectional study,

    C. S. Andreassen, M. D. Griffiths, R. Sinha, J. Hetland, and S. Pallesen, “The relationships between workaholism and symptoms of psychiatric disorders: A large-scale cross-sectional study,” PloS one, vol. 11, no. 5, p. e0152978, 2016

    2016

  10. [18]

    Employment and adults with autism spectrum disorders: Challenges and strategies for success,

    D. Hendricks, “Employment and adults with autism spectrum disorders: Challenges and strategies for success,” Journal of vocational rehabili- tation, vol. 32, no. 2, pp. 125–134, 2010

    2010

  11. [19]

    Factors influencing work participation of adults with developmental dyslexia: a systematic review,

    J. De Beer, J. Engels, Y . Heerkens, and J. van der Klink, “Factors influencing work participation of adults with developmental dyslexia: a systematic review,” BMC public health , vol. 14, pp. 1–22, 2014

    2014

  12. [20]

    Confidence in programming skills: Gender insights from stackoverflow developers survey,

    K. Kohl Silveira, S. Musse, I. H. Manssour, R. Vieira, and R. Prik- ladnicki, “Confidence in programming skills: Gender insights from stackoverflow developers survey,” in2019 IEEE/ACM 41st International Conference on Software Engineering: Companion Proceedings (ICSE- Companion...

    2019

  13. [21]

    What are developers talking about? an analysis of topics and trends in stack overflow,

    A. Barua, S. W. Thomas, and A. E. Hassan, “What are developers talking about? an analysis of topics and trends in stack overflow,” Empirical software engineering, vol. 19, pp. 619–654, 2014

    2014

  14. [22]

    Field, J

    A. Field, J. Miles, and Z. Field, Discovering Statistics using R . Sage, 2012

    2012

  15. [23]

    A simple sequentially rejective multiple test procedure,

    S. Holm, “A simple sequentially rejective multiple test procedure,” Scandinavian journal of statistics , pp. 65–70, 1979

    1979

  16. [24]

    Neurodiversity paradigms and their development across cultures: Some reflections in east asian contexts,

    T. Hirota, K.-A. Cheon, and M.-C. Lai, “Neurodiversity paradigms and their development across cultures: Some reflections in east asian contexts,” Autism, vol. 28, no. 11, pp. 2685–2689, 2024

    2024

  17. [25]

    Pandemic programming: How covid-19 affects software developers and how their organizations can help,

    P. Ralph, S. Baltes, G. Adisaputri, R. Torkar, V . Kovalenko, M. Kali- nowski, N. Novielli, S. Yoo, X. Devroey, X. Tan et al. , “Pandemic programming: How covid-19 affects software developers and how their organizations can help,” Empirical software engineering , vol. 25, pp. ...

    2020

  18. [26]

    Performance in inter- cultural interactions at work: Cross-cultural differences in response to behavioral mirroring

    J. Sanchez-Burks, C. A. Bartel, and S. Blount, “Performance in inter- cultural interactions at work: Cross-cultural differences in response to behavioral mirroring.” Journal of Applied Psychology , vol. 94, no. 1, p. 216, 2009

    2009

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.