About Prior Work
Free · Open access · No accounts · No data stored
What is this?
A free, open access psychosocial risk management tool built on complex data modelling techniques. The goal of Prior Work is to bring risk analytics - that once required specialist expertise - to any HR or WHS practitioner, with the inclusion of optional adaptive reasoning AI models to generate actionable, professional report summaries of existing workplace conditions.
A double meaning: the prior work your organisation has done - the survey, the data - is the foundation for analysis. The name also captures the core strength of the modelling approach itself - one built on incorporating new information to update previously held beliefs, and always being willing to update the world-view as fresh evidence arrives.
What this isn't
Not a clinical assessment tool
Group-level survey data only - no individual assessment, diagnosis, or monitoring. Clinical concerns need qualified practitioners.
Not proof of causation
The model surfaces statistical associations - factors that co-occur with elevated risk. Association is evidence, not causation; worker consultation explains the why.
Not a legal defence on its own
Consistent with the duty to identify psychosocial hazards, but not a substitute for the full risk management process required under WHS legislation.
Not a replacement for consultation
The data tells you where to look - not what you'll find or what to do. Structured worker consultation provides context and buy-in for change.
The methodology
Machine learning, not AI - structure learning and parameter estimation build models from your data. Results are mathematically computed, fully inspectable, and reproducible - unlike black-box AI.
Complex Data Modelling: leverages the interaction effects between hazards and outcomes as required by relevant WHS legislation.
Variable Elimination provides exact, deterministic inference - reproducible and instant regardless of model size.
Reverse propagation conditions on a target and compares how other factors change in turn. Large uplifts flag statistically associated factors - evidence to investigate, not proof of causation.
Privacy
Your data never leaves your browser session. The server computes on what you send and returns results immediately. Nothing is stored, logged, or associated with you - no accounts, no analytics on queries.
Uploaded survey CSVs are processed in memory for parameter fitting, then discarded. Only the fitted model (probability tables, not raw data) returns to your browser.
The optional Report generator uses the Claude API (your key) with your model output - not raw survey data. All analysis in Prior Work itself is deterministic.
Self-serve by design
No accounts, no data upload service, no subscription - by design. Your organisation uses its own survey platform, maintains ownership of its own data, and uses Prior Work as an analytical layer. The model lives in your browser - export it, share it, run it next year without depending on this service.
Where typical psychosocial platforms need long-term contracts, implementation projects, and data residency agreements, Prior Work is designed to be picked up and put down without a procurement cycle.
Use alongside other information
Findings are statistical patterns - a starting point, not a final answer. Output quality depends on data quality and representativeness. Prior Work surfaces uncertainty rather than hiding it. Results inform professional judgement and worker consultation; they don't replace them.
Treat this process as one layer - alongside worker consultation, observation, and other data. Default sector models are a starting point; collect your own organisation-specific data where available.
About the author
Patrick Egan is a registered, endorsed Organisational Psychologist specialising in Workplace Health and Safety. Connect with me on LinkedIn.
Developed with AI-assisted tooling (Claude, Anthropic) across architecture, front-end, analytics, and content. Analytical engine: pgmpy (FastAPI + pgmpy backend, React + Cytoscape.js front-end). Report generation uses the Claude API with your own key.
Grounded in decades of work stress research and the broader psychosocial risk science community. No affiliation with any regulatory body.
Cite Prior Work
If you use Prior Work in research, evaluation, or publication, please cite as:
Plain text
Egan, P. (2026). Prior Work: Psychosocial risk intelligence platform (Version 0.8) [Software]. Retrieved from https://priorwork.au.
BibTeX
@software{priorwork2026,
author = {Egan, Patrick},
title = {Prior Work: Psychosocial Risk Intelligence Platform},
year = {2026},
version = {0.8},
url = {https://priorwork.au},
}Contact & feedback
Bug reports, feature requests, research enquiries, accessibility issues:
patrick@priorwork.auBest-effort response times. Mark urgent accessibility issues with "Accessibility" in the subject line.