Automated Decision-Making
- Executive Summary
- Working Knowledge
- Technical Spec
GDPR Article 22 gives individuals the right not to be subject to decisions based solely on automated processing that produce legal or similarly significant effects. In the context of AI coaching, this means Sasha cannot make decisions about a person's development, career trajectory, or professional standing without meaningful human involvement. This is not a limitation — it is a design principle that makes the platform more trustworthy and more effective.
ReGenesis implements a clear separation of roles: Sasha suggests, humans decide. Every AI-generated insight, recommendation, and behavioral pattern assessment passes through a human-in-the-loop checkpoint before it influences coaching direction. The coach reviews and contextualizes Sasha's outputs, the coachee has full visibility into what the AI has inferred about them, and at no point does Sasha unilaterally take action that affects someone's professional life.
This design also addresses the broader concept of profiling — Sasha builds behavioral profiles to generate coaching insights, and individuals have the right to understand, challenge, and correct these profiles. Evidence Packs serve as the transparency mechanism, making Sasha's reasoning visible and contestable. This level of transparency is not just a legal requirement; it is what makes AI coaching credible.
GDPR Article 22: The Right to Human Involvement
What Article 22 Says
"The data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her."
What This Means for ReGenesis
| Concept | Application to ReGenesis |
|---|---|
| "Decision" | Any coaching recommendation, behavioral assessment, or developmental insight |
| "Solely automated" | Generated by Sasha without human review |
| "Profiling" | Sasha's behavioral pattern analysis and growth tracking |
| "Legal effects" | Unlikely in pure coaching context, but possible if coaching data influences employment decisions |
| "Similarly significant effects" | Coaching insights that shape someone's professional development and self-perception |
The Bottom Line
Sasha's insights are recommendations, not decisions. The coach is always in the loop. The coachee always has visibility. No automated process unilaterally determines someone's coaching direction.
Human-in-the-Loop Design
The Sasha → Coach → Coachee Flow
When Human Review Is Required
| Sasha Output Type | Direct to Coachee? | Coach Review Required? | Why |
|---|---|---|---|
| L0 Signal (brief observation) | Yes (if observe permission granted) | Not required but coach can see | Low-impact observation |
| L1 Summary (pattern analysis) | Yes | Coach notified | Moderate insight — coach should be aware |
| L2 Detail (full evidence) | Coach first | Required before sharing | High-impact insight with source evidence |
| Behavioral profile update | N/A (internal) | Coach can review | Profiling activity — transparency required |
| Coaching recommendation | No | Required | Directional advice — must be human-validated |
| Progress assessment | Yes (after coach review) | Required | Evaluative judgment — coach must contextualize |
| Nudge/prompt (act mode) | Yes (if act permission) | Coach can override | Proactive AI action — coach has override |
| Escalation trigger | Coach first | Required | Safety-critical decision |
Coach Override Capabilities
At any point, the coach can:
- Approve a Sasha insight — it becomes visible to the coachee
- Modify a Sasha insight — the coach edits the AI output before the coachee sees it (labeled as "Coach-modified AI insight")
- Reject a Sasha insight — it does not reach the coachee; coach provides reason for rejection
- Pause Sasha for a specific topic — if the AI is focusing on something the coach wants to handle differently
- Redirect Sasha's attention — coach can guide Sasha to focus on specific themes or patterns
- Escalate a Sasha concern — if the AI detects something the coach wants to discuss with the coachee in session
Human-in-the-loop is not just a legal requirement — it makes Sasha's outputs better. Coaches provide context that AI cannot access (body language, tone of voice, cultural context, client history), and their feedback improves Sasha's future outputs. The legal requirement and the product quality incentive are perfectly aligned.
Profiling Transparency
What Sasha Profiles
Sasha builds a behavioral profile of each coachee over time. This profile includes:
| Profile Element | Description | Example |
|---|---|---|
| Behavioral patterns | Recurring behaviors across sessions | "Avoids direct conflict 73% of opportunities" |
| Communication style | How the coachee communicates | "Tends toward analytical framing; less comfortable with emotional expression" |
| Growth areas | Identified development opportunities | "Decision-making under uncertainty" |
| Strengths | Observed strengths | "Strong strategic thinking; consistent follow-through" |
| Theme clusters | Topics that recur across sessions | "Leadership identity, work-life boundaries, team trust" |
| Sentiment trajectory | Emotional tone over time | "Overall positive trend with dips around board meetings" |
| Goal progress | Tracking against stated goals | "Goal 1: 60% progress; Goal 2: 30% progress" |
The Right to See Your AI Profile
Under GDPR (and good practice), coachees have the right to:
- See the behavioral profile Sasha has built about them
- Understand how each element was derived (via Evidence Packs)
- Challenge any element they believe is inaccurate
- Correct factual errors in the profile
- Request deletion of specific profile elements
- Opt out of profiling entirely (withdraw consent for Sasha's analyze mode)
How Challenges Work
- Coachee views their AI profile in the platform
- Each profile element has a "Challenge this insight" button
- Coachee provides reason for challenge (e.g., "This is based on a misunderstanding — I was role-playing a scenario, not describing my actual behavior")
- Challenge is routed to the coach for review
- Coach can: confirm the challenge (profile element removed/corrected), reject the challenge (with explanation), or escalate to governance review
- All challenges are logged and feed into bias monitoring
Exceptions to Article 22
Article 22 allows automated decisions in three cases:
| Exception | Description | ReGenesis Applicability |
|---|---|---|
| Contractual necessity | Automated decision is necessary for the contract | Limited — most coaching insights are not "necessary" in the strict sense |
| Authorized by law | Member state law allows it | May apply in some jurisdictions |
| Explicit consent | Data subject explicitly consents | Yes — but ReGenesis prefers human-in-the-loop over relying on this exception |
ReGenesis position: Even where exceptions apply, human-in-the-loop is maintained as a design choice. The exceptions exist as legal backstops, not as the primary operating model.
Sasha suggests, humans decide. This is not just a legal compliance statement — it is the product philosophy. The best coaching outcomes come from AI-augmented human judgment, not from AI replacing human judgment. Every technical decision should reinforce this principle.