PSP405TC · Ethics, Privacy & AI in Society
Fairness is not a property an algorithm possesses. It is a property a hiring system earns — through the rules it writes, the numbers it publishes, and the appeals it honours.
👆 Tap a CV — see what the machine sees
Bayesian reasoning shows how small early advantages compound: every positive signal raises the posterior probability of success — again and again, until early luck starts to look like talent.
A 10-year-old opens the posterior at 5%. Tap an early advantage to update the evidence.
No single signal decides anything. Together they compound — which is exactly why a screening model trained on "past successful hires" reads accumulated privilege as merit.
Canadian youth hockey groups children by calendar year, cut-off 1 January. At ten, eleven months is height, weight and strength. Scouts pick the visibly stronger — who get better coaching, more ice time, elite competition. The gap compounds until it looks like natural talent: the Matthew Effect, manufactured by a grouping rule.
The same logic runs inside résumé screening: "gap year" penalties, degree-tier filters and experience thresholds are all grouping rules — and the more “objective” they look, the easier it is to misread manufactured gaps as ability gaps.
In 1973, Berkeley's graduate school admitted 44% of men and 35% of women — and the cold aggregate nearly branded the university discriminatory. Broken down by department, most admitted women at equal or higher rates: women had simply applied to the most competitive departments. Simpson's Paradox.
One aggregate number, one accusation. Flip the view to see what it hid.
The numbers were never just numbers: applicants experience the aggregate as a verdict on their lives. Presentation is part of fairness — stratified rates, base rates, and plain-language context.
If the gap cannot be fully eliminated, governance must intervene — not with one grand fix, but at three complementary levels.
In 2024, a deaf Native American applicant alleged that HireVue garbled her answers, the employer refused her requested captioning, and the interview closed with an AI verdict to “improve communication skills”. The system did not merely miss her disability — it delivered a cold judgement that wounded her dignity.
Automated decisions are experienced as person-to-person verdicts. Warmth is not soft decoration.
This is not only fairness but human flourishing: giving those without early preparation a stage to shine, and protecting the vulnerable from careless harm. Progress is not only technological — it is also the inclusion between people.
Now turn the ideas into a tool. Upload your CV and our AI — running the exact fairness lenses from this page — will tell you how an automated screening system might misread it, and what to change before you hit “submit”.
Drop your CV here, or click to choose a file
PDF · DOCX · TXT · MD — parsed locally in your browser
Suggestions are generated by AI for education only — always verify against the actual job description and local law.
Write your story into the form — the preview updates as you type. Download it as an editable Word (.docx) file: the left side holds a photo slot, the right side stays blank for your name and contact details, so you decide what to reveal.
Everything happens in your browser — nothing is uploaded, nothing is stored.