BurnRiSc: Toward Non-Invasive Burnout Screening in Open Source from Public Repository Signals
Timofey Sanko, Yuan Tian, Mariam Guizani
cs.CY · cs.CL
Quick verdict
Claims GitHub activity traces can screen OSS burnout via an OLBI-based score; promising, but in-sample fitting and disclosure labels leave validity unproven.
Summary
BurnRiSc operationalizes the Oldenburg Burnout Inventory's exhaustion and disengagement dimensions as 14 behavioral and linguistic signals computed from GitHub activity (commits, pull requests, reviews, issue comments). Each signal is scored as a percentile level and four-month trend against the contributor's own history, smoothed, sigmoid-transformed, and aggregated into two weighted dimension scores averaged into a monthly Burnout Risk Score (BRS); weights are fit by L-BFGS-B to maximize median separation among 10 publicly self-disclosed burnout cases, 12 activity-collapse cases, and 46 comparison contributors across ten repositories. In the preliminary evaluation, sustained BRS elevation precedes 6 of 10 disclosures by 6–15 months; adding lifetime peak BRS as a second criterion raises coverage to 8 of 10 (5–16 months), and all 10 cases show at least one prior flag. Fitted weights are dominated by off-hours activity and commit volume, and peak BRS correlates only weakly with an independent activity score (r = 0.26). Main limitations: labels are keyword-found public disclosures rather than verified burnout; the threshold and peak criterion derive from the same fitted distributions used for evaluation, so headline coverage is largely in-sample; ten confirmed cases cannot identify fourteen weights; 11 of 46 comparison contributors were flagged with no outcome verification; and anonymized repositories limit exact replication. The authors position the work as a feasibility study ahead of a planned journal extension.
Review complete
Assessment at a glance
6
risk flags identified
● high 2
● medium 3
● low 1
5 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- OLBI's exhaustion and disengagement dimensions can be approximated by 14 behavioral and linguistic signals computed from public GitHub activity, scored against each contributor's own history.
- Sustained BRS elevation precedes 6 of 10 public burnout disclosures by 6-15 months; adding lifetime peak BRS raises coverage to 8 of 10 (5-16 months), and all 10 cases show at least one prior flag.
- Fitted weights are dominated by off-hours activity and commit volume, and peak BRS is not a proxy for contribution activity (r = 0.26).
- Burnout is therefore screenable from public repository data as an early-warning signal — screening, not diagnosis.
Strengths
Novel measurement contribution filling a real gap
First operationalization of a validated burnout instrument (OLBI) into repository-computable signals; §1-2 contrast this with CHAOSS's survey-only Project Burnout metric and the affect/toxicity literature that measures momentary state rather than chronic syndrome.
Self-referenced scoring design suited to OSS variance
§3.4 scores each signal as a percentile level plus 4-month trend against the contributor's own history, with a sigmoid offset to center 0.65 so that performing at one's historical norm does not register as elevated — a sensible answer to orders-of-magnitude volume differences across roles.
Unusual overfitting awareness for n=10
Purpose-built confound test (RQ3)
Candid limitations and deployment framing
Weaknesses
Label validity: disclosure is not burnout
Circular, in-sample evaluation of headline coverage
Severe sample limits and post-hoc anchor selection
Unverifiable comparison group and asymmetric precision argument
Thin text-signal validation and replication friction
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
BurnRiSc (Sanko, Tian, Guizani; Queen's University; JAWs '26 workshop, arXiv:2609.19422) claims the first operationalization of a validated burnout instrument — the two-dimensional Oldenburg Burnout Inventory (exhaustion, disengagement) — as 14 behavioral and linguistic signals computed from public GitHub activity, aggregated into a monthly Burnout Risk Score (BRS). The headline claim (Abstract, §5.1, Table 2): BRS elevation precedes public burnout disclosures — 6 of 10 cases by 6-15 months under a sustained-onset criterion, 8 of 10 with lifetime peak BRS as a second criterion, and 10 of 10 over any prior time frame. The authors frame this explicitly as a feasibility study for a planned TSE/TOSEM extension.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 5 weaknesses, analyzed in detail
- 6 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is BurnRiSc?
BurnRiSc is a framework that estimates a monthly burnout risk score (0-1) for open source contributors from public GitHub activity alone. It approximates the Oldenburg Burnout Inventory's two dimensions — exhaustion and disengagement — using 14 behavioral and linguistic signals (timestamps, throughput, commit/PR/review/comment text), and is explicitly positioned as a screening tool, not a diagnosis.
How well does it detect burnout?
In a preliminary study of 68 contributors including 10 people who publicly disclosed burnout, sustained score elevation preceded 6 of 10 disclosures by 6-15 months, and 8 of 10 when lifetime peak score counted as a second criterion. However, these results are computed in-sample (threshold and weights fit on the same data) and the labels are public disclosures rather than clinically verified burnout, so real-world accuracy is unestablished.
Is using this on contributors an ethics or privacy problem?
The paper argues a flag should trigger only a cheap, non-judgmental check-in, never an evaluation or intervention initiated by someone else, since the score could otherwise be gamed or weaponized. The authors withhold contributor identities but acknowledge that consent and appropriate-use questions for inferring psychological state from public traces remain to be worked out.
Is the code and data available?
Yes — the implementation and a supplementary document are at github.com/RISElabQueens/BurnRiSc. However, the ten repositories and all contributor identities are anonymized by number in the paper to protect the disclosers, so third parties cannot exactly re-derive the per-contributor results in Table 2 without the authors' replication package.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figure 1 (peak BRS vs activity score scatter) is referenced but its content cannot be inspected from extracted text.
- The supplementary document (per-anchor weight vectors, per-repository statistics, peak-risk distributions) was not part of the extracted text and carries part of the weight-stability evidence.
- The ACM reference block contains template placeholders (DOI/ISBN), consistent with a preprint; repository anonymization means Table 2 cannot be externally spot-checked.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.