Presentation on the IEEE Global Initiative for Ethics of Autonomous and Intelligent Systems presented at the KAIST workshop on Taming AI: Engineering, Ethics and Policy, June 2018

1. 1
IEEE Global Initiative on Ethics of
Autonomous and Intelligent Systems:
Industry Standards and Ethically Aligned Design
Ansgar Koene, Chair P7003 Algorithmic Bias Considerations
Senior Research Fellow, University of Nottingham
21 June 2018

2. 2

3. 3

4. 4

5. Ethically Aligned Design, v2
5
• More than one hundred pragmatic recommendations for
technologists, policy makers and academics
• Created by 250+ global cross-disciplinary thought leaders

6. Summary of content EAD v2
Committees featured in EADv1 (with updated content)
 General Principles 20-32
 Embedding Values into Autonomous Intelligent Systems 33-54
 Methodologies to Guide Ethical Research and Design 55-72
 Safety and Beneficence of Artificial General Intelligence
(AGI) and Artificial Superintelligence (ASI) 73-82
 Personal Data and Individual Access Control 83-112
 Reframing Autonomous Weapons Systems 113-130
 Economics/Humanitarian Issues 131-145
 Law 146-161
New Committees for EADv2 (with new content)
 Affective Computing 162-181
 Policy 182-192
 Classical Ethics in A/IS 193-216
 Mixed Reality in ICT 217-239
 Well-being 240-263
6

7. IEEE P70xx Standards Projects
IEEE P7000: Model Process for Addressing Ethical Concerns During System Design
IEEE P7001: Transparency of Autonomous Systems
IEEE P7002: Data Privacy Process
IEEE P7003: Algorithmic Bias Considerations
IEEE P7004: Child and Student Data Governance
IEEE P7005: Employer Data Governance
IEEE P7006: Personal Data AI Agent Working Group
IEEE P7007: Ontological Standard for Ethically Driven Robotics and Automation
Systems
IEEE P7008: Ethically Driven Nudging for Robotic, Intelligent and Autonomous
Systems
IEEE P7009: Fail-Safe Design of Autonomous and Semi-Autonomous Systems
IEEE P7010: Wellbeing Metrics Standard for Ethical AI and Autonomous Systems
IEEE P7011: Process of Identifying and Rating the Trustworthiness of News Sources
IEEE P7012: Standard for Machines Readable Personal Privacy Terms
7

8. 8
http://sites.ieee.org/sagroups-7003/

9. Algorithmic Discrimination
9

10. Confirmation Bias: Predictive policing
10
 A need for counterfactual double blind trials

11. Complex individuals reduced to
simplistic binary stereotypes
11

12. Case study: Recidivism risk prediction
 COMPAS recidivism prediction tool
– Built by a commercial company, Northpointe, Inc.
 Estimates likelihood of criminals re-offending in future
– Inputs: Based on a long questionnaire
– Outputs: Used across US by judges and parole officers
 Are COMPAS’ estimates fair to salient social groups?
12
Machine Bias: There’s software used across the
country to predict future criminals. Propublica

13. Case study: Recidivism risk prediction
13
Is the algorithm fair
to all groups?
When base rates differ, no non-trivial solution can achieve similar FPR,
FNR, FDR, FOR!

14. 14
Open invitation to join the P7003 working group
http://sites.ieee.org/sagroups-7003/

15. Key question when developing or
deploying an algorithmic system
15
 Who will be affected?
 What are the decision/optimization criteria?
 How are these criteria justified?
 Are these justifications acceptable in the context where the system is
used?

16. P7003 foundational sections
 Taxonomy of Algorithmic Bias
 Legal frameworks related to Bias
 Psychology of Bias
 Cultural aspects
16
P7003 algorithm development sections
 Algorithmic system design stages
 Person categorization and identifying affected population groups
 Assurance of representativeness of testing/training/validation data
 Evaluation of system outcomes
 Evaluation of algorithmic processing
 Assessment of resilience against external manipulation to Bias
 Documentation of criteria, scope and justifications of choices

17. Related AI standards activities
 British Standards Institute (BSI) – BS 8611 Ethics design and application
of robots
 ISO/IEC JTC 1/SC 42 Artificial Intelligence
– SG 1 Computational approaches and characteristics of AI
systems
– SG 2 Trustworthiness
– SG 3 Use cases and applications
– WG 1 Foundational standards
 Jan 2018 China published “Artificial Intelligence Standardization White
Paper.”

18. ACM Principles on Algorithmic
Transparency and Accountability
 Awareness
 Access and Redress
 Accountability
 Explanation
 Data Provenance
 Auditability
 Validation and Testing
18

19. FAT/ML: Principles for Accountable
Algorithms and a Social Impact
Statement for Algorithms
 Responsibility: Externally visible avenues of redress for adverse effects, and
designate an internal role responsible for timely remedy of such issues.
 Explainability: Ensure algorithmic decisions and data driving those decisions
can be explained to end-users/stakeholders in non-technical terms.
 Accuracy: Identify, log, and articulate sources of error and uncertainty so that
expected/worst case implications can inform mitigation procedures.
 Auditability: Enable third parties to probe, understand, and review algorithm
behavior through disclosure of information that enables monitoring, checking,
or criticism, including detailed documentation, technically suitable APIs, and
permissive terms of use.
 Fairness: Ensure algorithmic decisions do not create discriminatory or unjust
impacts when comparing across different demographics.
https://www.fatml.org/resources/principles-for-accountable-algorithms
19

20. Thank you for your attention
20
http://unbias.wp.horizon.ac.uk/
ansgar.koene@nottingham.ac.uk