COMP90087 overview Tim Miller, Kobi Leins, Simon Coghlan, Marc Cheong School of Computing and Information Systems Centre for AI & Digital Ethics The University of Melbourne 3Subject overview Who are we? 4 Simon Kobi Tim Marc Tutors 5 Kate Michael Matt COMP90087 Learning outcomes On completion of this subject, students should: • Understand and describe the social and ethical issues raised by technology in various contexts • Apply established ethical theories to reason about social and ethical issues related to technology • Critically examine, analyse, and apply ethical perspectives to design, develop and deploy of digital technologies responsibly • Communicate perspectives to other technical professionals and the community at large • Demonstrate a profound respect for that fact that just because you can, does not mean you should create a technology 6 COMP90087 Assessment 7 Description Timing Percentage Tutorial participation and contributions to online discussion. Two hours per week. Intended Learning Outcomes (ILO's) 1 to 4 are addressed by tutorial participation and contributions to online discussion. 2% each; best 10 of 12 Throughout the teaching period 20% Critical evaluation essay. ILO's 1 to 5 are addressed in the essay. • 35-45 hours Week 7 30% Research essay. Due first week of examination period. ILO's 1 to 5 are addressed in the essay. • 35-45 hours During the examination period 30% Online quiz. ILO's 1 and 2 are addressed in the online quiz. • 1 hours End of semester 20% COMP90087 Agenda One module per week Each module: • Readings (1-2 papers) • Video playlist (30-60 minutes), with notes • Consultation & discussion hour: 9-10am on Thursdays (on Wednesday in week 2) • Tutorial Group discussion on relevant topic 8 Interaction When you need advice, input, direction: • Check Canvas for announcements. • Check the relevant discussion forum for existing answers. • Post a question to the relevant discussion forum Feel free to answer others’ questions! • Attend a consultation hour (Thu 9am Melbourne time). • Ask your tutor during the tutorial. • For personal matters, send an email: Please do not email more than one staff member independently about the same issue. It duplicates work between us. • We commit to responding within 1 working day Please do not expect replies over the weekend 9 10 The Ethics of Artificial Intelligence Six types of ethical concerns 11 Inscrutable evidence Inconclusive evidence Traceability Transformative effects Unfair outcomes Misguided evidence Epistemic concerns Normative concerns Tsamados, Andreas, et al. "The ethics of algorithms: key problems and solutions." Available at SSRN 3662302 (2020). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3662302 Six types of ethical concerns 12 Inscrutable evidence Inconclusive evidence Traceability Transformative effects Unfair outcomes Misguided evidence Caused by: • Incomplete data sets • Data sets with structural inequalities Impact: • Erroneous actionable insights Six types of ethical concerns 13 Inscrutable evidence Inconclusive evidence Traceability Transformative effects Unfair outcomes Misguided evidence Caused by: • Lack of appropriate tools for transparency • Cognitive impossibility of interpretation Impact: • Erroneous actionable insights • Shirk responsbility Six types of ethical concerns 14 Inscrutable evidence Traceability Transformative effects Unfair outcomes Inconclusive evidence Misguided evidence Caused by: • Use of unsuitable features/variables • Use of model in different context Impact: • Biased algorithms Six types of ethical concerns 15 Inscrutable evidence Traceability Transformative effects Misguided evidence Inconclusive evidence Unfair outcomes Caused by: • Inconclusive or misguided evidence Mitigations • Third-party auditing • Collaborative community-based design Six types of ethical concerns 16 Inscrutable evidence Traceability Unfair outcomes Misguided evidence Inconclusive evidence Transformative effects Caused by: • Users’ limited understanding of algorithms • Choices constrained by algorithms • Insufficient recourse Impact: • Reduction in human autonomy Six types of ethical concerns 17 Inscrutable evidence Transformative effects Unfair outcomes Misguided evidence Inconclusive evidence Traceability Caused by: • Lack of transparency, explainability, and accountability Impacts: • Lack of moral responsbility • “Agency laundering” 18 Syllabus COMP90087 Syllabus 19 Week Module Coordinator 1 Trust, machine, and digital ethics Tim 2 The History of Artificial Intelligence Tim 3 Philosophy and ethics Simon 4 Fairness and accountability Simon 5 Data governance Marc 6 Accessibility and equity Marc 7 Transparency: Decisions & processes Marc 8 Explainability Tim 9 Politics and policy Kobi 10 Frameworks and implementation Kobi 11 Reliability and safety Kobi 12 Bringing it together Simon Have fun!
欢迎咨询51作业君