AI & Gig-Work Fairness Policy

1. Purpose

To ensure Able's AI systems and marketplace practices embody our core values: Be Good • Be Fair • Be Kind.

2. AI governance

Transparency	Plain-English summary of matching algorithm factors is published (weights table). Users can request additional explanation of any automated recommendation.
Contestability	Workers or buyers may trigger human review of a match or decision within 72 hours.
Bias audit	Independent audit of matching scores every quarter from Oct 2025; headline metrics published on our blog.
Data minimisation	Only shift-relevant attributes (skills, availability, location radius < 15 mi) are passed into the model. Sensitive traits (age, gender, ethnicity) are not captured by Able matching agent.
Safety & security	Model inputs/outputs logged, salted and retained 12 months for incident response; penetration testing monthly.

3. Fair-work standards

Earnings	Minimum rate ≥ statutory National Minimum/Living Wage for the role; same-day payout (Stripe Express) at 100% of agreed hours.
Autonomy	Freelancers set their own rates, availability and max travel distance; no exclusivity clauses.
Transparency	All fees displayed to both parties before booking; Able's 8% + VAT buyer fee is fixed and published.
Representation	Community channel plus quarterly round-tables with the CEO & Community Manager.
Well-being	System flags burnout risk (≥40 hrs in rolling 7 days); prompts rest and will throttle offers.
Dispute resolution	Escalation path: Chat → Community Manager → Independent Arbitrator (CEDR model rules) within 28 days.

4. Reputation & Feedback (No Ratings)

No public star-ratings	Able does not display 1-to-5 stars, streaks or productification reputation metrics for workers or buyers. We consider such numerical scoring systems to reinforce confirmation bias, cause stress and can entrench discrimination.
Context-rich feedback instead	After each gig both parties are invited to give structured feedback via short free-text prompts ("What went well?" / "What could be improved?"). This is visible only to the counter-party unless they chose to make it public and Able's Trust & Safety Team.
AI-assisted summarisation – human-reviewed	Where many gigs occur, a natural-language summary is produced to help users surface common themes; a human moderator checks for tone and inadvertent disclosure of personal data before release. These summaries can be contested via our discord channel and community manager.
Use of feedback in matching	Qualitative insights may be converted into a binary 'trusted partner' flag (visible only to Able's internal systems) after manual verification. The flag can increase the weighting of future matches but never blocks access to gigs; any negative pattern triggers human review.
Right to challenge	Workers and buyers can request deletion or redaction of feedback they consider unfair or inaccurate. Disputes follow the escalation path set out in §3 "Dispute resolution".

5. Monitoring & reporting

Key indicators (fill-rate, cancellation rate, wage compliance, bias metrics) are reported on a public Fairness Dashboard updated monthly.