CyberPsyche Logo

Ethical Impact Assessment

CyberPsyche · AI Ethics Framework

Welcome to the EIA Tool

The Ethical Impact Assessment (EIA) has been developed based on a number of existing frameworks and guidelines, benchmarked against other publicly available impact assessments. Use this tool to systematically evaluate the ethical implications of your AI system or campaign.

How Scoring Works

Score the potential negative impact that a campaign or engagement may have. Rate how severe the impact would be to the targeted user/audience, then score how large the impact would be in terms of scale of campaign/engagement (number of people affected). These scores are multiplied and totalled resulting in an Impact score per principle. These are further calculated to produce an overall impact value and required action.

High risk AI systems — such as those used in employment, financial and health outcomes — can have severe impact on users if ethical principles are breached.

Impact Rating Levels & Required Actions

< 1
Minimal ethical impact. Continue current deployment with annual reviews. Document ethical rationale in campaign plans for transparency.
< 5
Low level of ethical impact. Enhance documentation and conduct quarterly spot-checks. Train teams on AI ethics.
< 10
Medium level of ethical impact. Implement immediate low-effort remediations. Develop a 6-month action plan with metrics for progress tracking.
> 10
High level of ethical impact. Pause high-risk activities, perform full audits, and engage ethics/legal experts. Update policies and retrain all staff within 30 days.
> 15
Extreme level of ethical impact. Halt deployment in the affected area, notify regulators/stakeholders, and execute a root-cause analysis.
📊
Scoring Guidelines
Use these scales when rating severity and scale
Severity (1–5)
1Minimal harm; fully reversible, affects few stakeholders.
2Minor, localised impact; easily mitigated, short-term effects.
3Noticeable harm to some users/groups; requires mitigation.
4Significant, widespread harm; difficult to reverse.
5Severe, irreversible damage; fundamental rights violation.
Scale (1–5)
1Affects <1% of users or single instances.
2Impacts small group or localised scope (<10%).
3Affects substantial subset or national scope (10–50%).
4Widespread impact across markets/users (>50%).
5Universal or societal-level reach; all users/market.

Frameworks & Guidelines

OECD
AI Principles 1.1–1.5oecd.org/en/topics/sub-issues/ai-principles
GDPR
Chapter 2, Article 5gdpr.eu/tag/chapter-2/
UNESCO
Ethical Impact Assessmentunesco.org/ethics-ai/en/eia
Public Voice Coalition
Universal Guidelines for AI (UGAI)caidp.org/universal-guidelines-for-ai/
EU AI Act
EU Artificial Intelligence Acteur-lex.europa.eu
African Union
Continental AI Strategyau.int
Council of Europe
Framework Convention on AI, Human Rights, Democracy & Rule of Lawrm.coe.int/1680afae3c
ISO 42001
Artificial Intelligence Management System (AIMS)iso.org/standard/42001
📋
Section 1 — Project Overview
Provide details about the AI system or campaign being assessed
Attach governance structure if available
⚠ Risk if unclear: Proportionality not established
✅
Section 1 — Project Criteria
Answer the following questions about your AI system
⚠ Risk: Data breaches
⚠ Risk: Scale of impact (scope & timescale)
⚠ Risk: Remediation required
⚠ Risk: Inherent biases in the training model
If yes, attach outcomes
⚠ Risk: Data breaches, errors
If yes, attach report
⚠ Risk: Biased outcomes
If yes, attach plan
⚠ Risk: Biased outputs, lack of inclusivity
If yes, attach assessment
⚠ Risk: Not compliant with sustainability reporting
If yes, attach assessment
⚠ Risk: Data breaches
⚠ Risk: Inequality, discrimination, harm
⚠ Risk: Lack of understanding of ethical implications
⚖
Section 2 — Ethical Impact Assessment
Rate severity (1–5) and scale (1–5). Impact Score = Severity × Scale.
1. AutonomyScore: 0
Rate if Yes — Manipulation, deception or dark patterns are used:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Autonomy Impact Score0
2. OptionalityScore: 0
Rate if No — Scoring reflects absence of protective measures:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Optionality Impact Score0
3. Privacy & Data ProtectionScore: 0
Personal data is used — Rate if Yes:
Data Type Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Measures taken to protect personal data — Rate if No:
Measure Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Privacy & Data Protection Impact Score0
4. Fairness, Bias & InclusionScore: 0
Rate if No — Bias testing conducted:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Rate if No — Bias mitigation adopted:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Fairness, Bias & Inclusion Impact Score0
5. Transparency & ExplainabilityScore: 0
Rate if No — Transparency measures are in place:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Transparency Impact Score0
6. AccountabilityScore: 0
Rate if No — Absence of accountability measures:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Accountability Impact Score0
7. SustainabilityScore: 0
Rate if Yes — Environmental impact concerns are present:
Criterion Y / N Severity (1–5) Scale (1–5) Impact Priority Proposed Mitigation Owner Status
Sustainability Impact Score0
-
Total Impact Score
-
Average per Principle
-
Risk Level

Recommended Action

Score by Principle