Certainty is another factor. In any case, I agree with @David_Gunkel that conversations should include alternative ethical theories. A lot of the talking past each other is based in philosophical issues that have been studied for a long time, but rarely in CS curricula!
@petitegeek
-
Utilitarian Calculation Limits: Temporal Weighting in Decision Making
By
–
One of the major drawbacks of utilitarianism is that it is hard to draw the line for when to stop the calculation. What to include? A scenario millions of years from now? According to Bentham, we should consider benefits happening sooner with more weight. https://
en.m.wikipedia.org/wiki/Felicific
_calculus
… -
Kantian Ethics and Rights-Based Language in AI Governance
By
–
But also has Kantian language such as "people shouldn't be used as a means to an end" and rights-based language (consent, inclusivity, data access) that can be used in concert with the utilitarian calculus.
-
AI Ethics Principles Combining Utilitarianism and Cost-Benefit Analysis
By
–
One example of a set of AI principles that contains a mix of ethical perspectives: https://
thehumeinitiative.org/guiding-princi
ples/
… It contains utilitarianism "benefits must outweigh the harms" and other principles that help support the accurate calculation of costs/benefits -
Social Contract and Human Rights in AI Ethics
By
–
Social contract theory talks about rights, like all of those written in the UN Declaration of Human Rights.
-
Kantian Ethics and AI: Non-harm and Human Dignity Principles
By
–
For example, Kantianists argue that we should consider rules such as "you should not kill", and that we shouldn't use people as a means to an end.
-
Four Ethical Frameworks for Technology Ethics Courses
By
–
In my course on tech ethics, we use the textbook Ethics for the Information Age that considers 4 "workable" ethical theories: act/rule utilitarianism, Kantianism (deontology), social contract theory and virtue ethics.
-
Training Data Bias: Gender Representation in AI Image Datasets
By
–
Could there be anything accounting for general exposure to images of people in daily life? E.g. magazines for women depicting women, magazines for men depicting women. Or forward facing images with eye contact?
-
Algorithm Reversibility: Control in Autonomous Decision-Making Systems
By
–
Is the decision your algorithm makes reversible? e.g. light switch vs potentially lethal autonomous car. This can impact how much control a user feels they have. Insights from this great talk on causality, from a philosopher of neuroscience.
-
Understanding SOTA Models: Scientific Gaps and Research Impact
By
–
My guess: lack of scientific understanding of how, when and why the SOTA models work (or don't). That just data+++ can improve a system is also unsatisfying. Finally they'd rather focus on research that have a clear connection to improving the world (e.g. brain science).