Pierre Stein­er, Direc­tor of UTC’s TSH (Tech­nol­o­gy and Social Sci­ences) depart­ment, teach­es phi­los­o­phy and epis­te­mol­o­gy. Among the sub­jects he teach­es is ethics and the rela­tion­ship between ethics, tech­nol­o­gy and the engi­neer­ing profession.

This nat­u­ral­ly makes AI tech­nolo­gies a field for study and reflec­tion, in order to mea­sure their social and envi­ron­men­tal con­se­quences, for exam­ple. “Today, the rela­tion­ship between ethics and AI is expressed in dif­fer­ent ways and these vari­a­tions are not nec­es­sar­i­ly a func­tion of the types of AI that exist. If we take the most recent gen­er­a­tive AIs, their lack of trans­paren­cy has both tech­no­log­i­cal dimen­sions for design­ers, but also eth­i­cal and legal lev­els. Among eth­i­cal val­ues, we might ask whether these sys­tems respect, for exam­ple, fair­ness or the absence of bias. We can also raise the ques­tion of respon­si­bil­i­ty — is it the sys­tem that is respon­si­ble, or the sys­tem design­er? We can also look at the way in which the data has been col­lect­ed, to ensure that it reflects the diver­si­ty of the pop­u­la­tion and like­wise for pri­va­cy issues. AI also rais­es new legal issues. If we don’t know how the sys­tem goes about pro­duc­ing this or that response, how can we be sure that a response is jus­ti­fied? And how can we allow such a non-trans­par­ent sys­tem to sup­port deci­sions nor­mal­ly tak­en by humans?

Indeed, while the ben­e­fits derived from AI data and algo­rithms may be sig­nif­i­cant, the risks of error do exist and are unac­cept­able in a large num­ber of fields. “If we take the health­care sys­tem, for exam­ple, would we want to del­e­gate deci­sion-mak­ing about a diag­no­sis or the care to be pro­vid­ed to these sys­tems? How far are we pre­pared to go and at what point do we feel that human inter­ven­tion is still nec­es­sary? But more gen­er­al­ly — and this is an emi­nent­ly polit­i­cal issue — we need to ask our­selves what kind of Soci­ety and world we want to shape. Do we want a world in which AI plays a greater role in the way we are informed, pro­duce knowl­edge, teach or work? We may also won­der about the con­se­quences of AI in soci­etal terms, with the increas­ing automa­tion of many tasks, but also in eco­log­i­cal terms. In short, we need to make choic­es and define the place we want AI to occu­py in Soci­ety. These are not just ques­tions for engi­neers and design­ers. They are polit­i­cal and eth­i­cal ques­tions that can­not be resolved by algo­rithms or data aggre­ga­tion”, he asserts.

These are eth­i­cal ques­tions that per­me­ate teach­ing and research at UTC. “What’s spe­cial about the UTC is that we don’t think about eth­i­cal issues at the end of the chain, when the sys­tem is deployed. They have to be addressed upstream. It’s right from the design stage of the sys­tem that you have to ask your­self what’s at stake, from an eth­i­cal point of view,”, con­cludes Pierre Steiner.

MSD