“Data is an asset but is it also a risk” (Hasselbach and Tranberg, 2017)
All data that is collected and processed has an impact on people in some way and has the potential to create great benefit but can also cause harm if not properly handled (O’Keefe and O’Brien, 2018). The potential for benefit and harm is amplified when we use that data to create AI systems intended to make decisions or predictions on our behalf.
In order to design human centred AI systems, we need to understand the issues of inequalities, bias and discrimination that exist in our society and can be perpetuated in our data and data driven systems. We need to understand and adhere to legal obligations with regards to data protection and management of personal information (i.e. GDPR in the EU). It is also important to look beyond data specific legislation to existing laws, policies and social theories that promote equality and protect against bias and discrimination and are central to a human centred design approach.
We also need to recognise that while technology advances at a rapid pace across territories and boundaries, the law moves notably slower and is usually created within physical regional boundaries. The field of ethics provides useful concepts and frameworks to identify risk and understand best practices for the design of human centred AI that can extend beyond basic legal obligations or replace the law where it does not yet exist. This is particularly important when we consider “Future AI” and propose novel innovative AI systems, we need to keep ethics, equality and human rights to the forefront of our design.
Week 1-2: Society and AI
- Examples of challenges and opportunities for AI systems explored through an ethical and civic lens using case studies in domains such as access to credit, healthcare, employment and criminal justice.
- Challenges with datasets and discrimination - lack of diversity in existing data sets, selection bias in new data collection
- Challenges with models and discrimination - Poorly designed matching systems, Personalization and recommendation services that narrow user options and perpetuate discrimination
Week 3: EU and International legislation and frameworks on data, AI, human rights and equality
- Data – (EU GDPR, US COPPA, HIPPA)
- AI – ( EU Proposal for Regulation on Artificial Intelligence)
- EU Human Rights Legislation (Irish Human Rights and Equality Commission Act 2014)
- Equality - Equal Status Acts 2000–2018 (the ‘Acts’); European Union (Accessibility of Websites and Mobile Applications of Public Sector Bodies) Regulations 2020
- Explore strengths and limitations of existing laws
Weeks 4-5: Data Management
- Data roles and responsibilities
- Governance and Stewardship
- Key data stakeholders
- Data Protection and privacy
- Managing Personal Data - anonymisation and pseudonymisation
- Data Management Plans (in academia and industry)
Weeks 6-7: Audits and Assessment
- Data Protection
- Security
- Safety
- Equality and Inclusion
Weeks 8-9: Policy and Frameworks to Preventing Discriminatory Outcomes in AI Systems based on (WEF, 2016) framework
- Active Inclusion
- Fairness
- Right to Understanding
- Access to Remedy
Weeks 9-10: Human Centred Data Lifecycle – Ethics, risk and mitigation
- Explore each step of the AI/ML lifecycle using for example CRISP-DM from a human centred perspective. For example during a data collection phase, what are the risks for excluding participants from minorities (age, gender, disability, race) and how can we ensure inclusive study design and recruitment practices?
Week 11-12: Future of AI
- Case studies and test cases that go beyond current legislation
- Ethical tools and frameworks to scaffold novel systems where relevant legislation does not exist.
This module will be presented over 13 lectures, and copies of the lecture material will be provided. Students will be expected to use library and internet based information sources extensively, and familiarise themselves with the supplemental reading. Students will be encouraged to be proactive in their approach to learning for this module which will require independent research and critical thinking. The continuous assessment will take the form of case studies, reports and presentations. Students will be required to work both independently and as part of groups. Where possible real-world examples will be used.
| Module Content & Assessment | |
|---|---|
| Assessment Breakdown | % |
| Other Assessment(s) | 100 |