These dis\u00adtor\u00adtions and hal\u00adlu\u00adci\u00adna\u00adtions<\/strong> are usu\u00adally dif\u00adfi\u00adcult for users to rec\u00adog\u00adnize if AI providers do not make the train\u00ading data sets used and the func\u00adtion\u00ading of the machine learn\u00ading algo\u00adrithms used trans\u00adpar\u00adent. <\/li>\n<\/ul>\n\n<\/p>\n\n
Users should always ask them\u00adselves whether AI results stand up to their own scrutiny and whether their dis\u00adsem\u00adi\u00adna\u00adtion respects the equal oppor\u00adtu\u00adni\u00adties and dig\u00adnity of all peo\u00adple!<\/strong><\/p>\n\n\n- Users do not only oper\u00adate AI tools with their user input (so-called \u201cprompts\u201d). This user input can also be used as train\u00ading data for the AI. If users enter their per\u00adsonal data or the per\u00adsonal data of other peo\u00adple, this may not only vio\u00adlate data pro\u00adtec\u00adtion<\/strong> reg\u00adu\u00adla\u00adtions. Espe\u00adcially if AI sys\u00adtems are fed with unrec\u00adog\u00adnized dis\u00adcrim\u00adi\u00adna\u00adtory prej\u00adu\u00addices, these could be applied to per\u00adsonal data and dis\u00adsem\u00adi\u00adnated. Uni\u00adver\u00adsi\u00adties should not only develop clear guide\u00adlines for the han\u00addling of sen\u00adsi\u00adtive data when using AI, but also estab\u00adlish pro\u00adce\u00addures<\/strong> with which teach\u00aders and stu\u00addents can for\u00admally approve the use of their own mate\u00adri\u00adals \u2014 such as prompts, tem\u00adplates or assign\u00adment for\u00admats \u2014 by oth\u00aders. Tem\u00adplates for the trans\u00adfer of use can be pro\u00advided for this pur\u00adpose, e.g. to allow third par\u00adties to reuse con\u00adtent in AI sys\u00adtems. <\/li>\n<\/ul>\n\n
Users should always ask them\u00adselves whether enter\u00ading per\u00adsonal data could harm them or oth\u00aders, whether the data can be anonymized or whether it can be dis\u00adpensed with alto\u00adgether!<\/strong><\/p>\n\n\n- When using AI sys\u00adtems, users may del\u00ade\u00adgate deci\u00adsions to the AI. Espe\u00adcially when such deci\u00adsions affect or could affect other peo\u00adple, there is a risk that the use of AI in deci\u00adsion-mak\u00ading processes, for exam\u00adple in selec\u00adtion pro\u00adce\u00addures, objec\u00adti\u00adfies<\/strong> those peo\u00adple<\/strong>. This means that deci\u00adsions are made over their heads. Deci\u00adsions should not be made with\u00adout involv\u00ading those (poten\u00adtially) affected in the deci\u00adsion-mak\u00ading process and the rea\u00adson\u00ading, as oth\u00ader\u00adwise their human dig\u00adnity may be vio\u00adlated. <\/li>\n\n\n\n
- Fur\u00adther\u00admore, when AI sys\u00adtems are used in deci\u00adsions, it is not always clear who is respon\u00adsi\u00adble<\/strong> for those deci\u00adsions, which can lead to fur\u00adther ques\u00adtions of lia\u00adbil\u00adity. Uni\u00adver\u00adsi\u00adties are par\u00adtic\u00adu\u00adlarly chal\u00adlenged here to define clear respon\u00adsi\u00adbil\u00adi\u00adties \u2014 both in admin\u00adis\u00adtra\u00adtive and teach\u00ading oper\u00ada\u00adtions. <\/li>\n<\/ul>\n\n
Users should ask them\u00adselves whether they should leave final deci\u00adsions to AI sys\u00adtems. Mak\u00ading final deci\u00adsions your\u00adself means tak\u00ading respon\u00adsi\u00adbil\u00adity for your own actions, which may affect other peo\u00adple, and respect\u00ading the dig\u00adnity of oth\u00aders! <\/strong><\/p>\n\n5 principles for dealing with AI at universities<\/h2>\n\n
Uni\u00adver\u00adsi\u00adties can meet these and other chal\u00adlenges in deal\u00ading with AI by devel\u00adop\u00ading and defin\u00ading bind\u00ading rules for deal\u00ading with AI in uni\u00adver\u00adsity teach\u00ading. How\u00adever, teach\u00aders and stu\u00addents should also strive for their own eth\u00adi\u00adcal debate on AI. In ethics, jus\u00adti\u00adfi\u00adca\u00adtions for morally good actions are devel\u00adoped in response to the ques\u00adtion \u2018What should I do? To this end, the inten\u00adtions for actions, the actions them\u00adselves and their pos\u00adsi\u00adble con\u00adse\u00adquences, such as ben\u00ade\u00adfits and harms for one\u00adself and oth\u00aders, are exam\u00adined and weighed up.<\/p>\n\n
As an exam\u00adple, five clear prin\u00adci\u00adples for deal\u00ading with AI accord\u00ading to Floridi, et al., 2019 are pre\u00adsented here:<\/p>\n\n
\n- Benef\u00adi\u00adcence: AI should only be used in a way that pro\u00admotes human wel\u00adfare, pre\u00adserves human dig\u00adnity and is sus\u00adtain\u00adable.<\/li>\n\n\n\n
- Non-malef\u00adi\u00adcence: Espe\u00adcially when risks can\u00adnot be com\u00adpletely avoided or there is uncer\u00adtainty about the occur\u00adrence of pos\u00adsi\u00adble risks, a prin\u00adci\u00adple of dam\u00adage avoid\u00adance and dam\u00adage lim\u00adi\u00adta\u00adtion should also be pur\u00adsued.<\/li>\n\n\n\n
- Auton\u00adomy: The prin\u00adci\u00adple of auton\u00adomy calls on users to weigh up the trans\u00adfer of deci\u00adsion-mak\u00ading and deci\u00adsion-mak\u00ading pow\u00aders to AI against the preser\u00adva\u00adtion of their own and oth\u00aders\u2019 free\u00addom of action and deci\u00adsion-mak\u00ading. All pos\u00adsi\u00adble stake\u00adhold\u00aders must be included in this con\u00adsid\u00ader\u00ada\u00adtion and par\u00adtic\u00adu\u00adlar pri\u00ador\u00adity must be given to safe\u00adguard\u00ading human free\u00addom of action and deci\u00adsion-mak\u00ading and their dig\u00adnity. <\/li>\n\n\n\n
- Jus\u00adtice: The use of AI should serve to pro\u00admote the well-being of every human being, to main\u00adtain sol\u00adi\u00addar\u00adity, and to avoid injus\u00adtice. When deal\u00ading with AI, the pos\u00adsi\u00adble occur\u00adrence of dis\u00adcrim\u00adi\u00adna\u00adtion (e.g. in the con\u00adclu\u00adsions pro\u00advided or in the under\u00adly\u00ading data sets used by an AI) should be taken into account, bal\u00adanced or at least mit\u00adi\u00adgated. <\/li>\n\n\n\n
- Explic\u00ada\u00adbil\u00adity: The use of AI obliges \u2014 as far as pos\u00adsi\u00adble \u2014 to make one\u2019s own usage inten\u00adtions under\u00adstand\u00adable and to accept one\u2019s own account\u00adabil\u00adity for the result\u00ading con\u00adse\u00adquences of use. In addi\u00adtion, these aspects must also be made trans\u00adpar\u00adent so that the ben\u00ade\u00adfits, risks and dam\u00adage can be dis\u00adcussed by all those involved and by soci\u00adety as a whole. <\/li>\n<\/ol>\n\n
And now? What should be done in this spe\u00adcific case? <\/strong><\/p>\n\nEthics can\u00adnot (and should not) pro\u00advide rigid guide\u00adlines or instruc\u00adtions to be fol\u00adlowed. Instead, applied ethics thrives on our own seri\u00adous engage\u00adment with new and exist\u00ading chal\u00adlenges and the open and trans\u00adpar\u00adent exam\u00adi\u00adna\u00adtion of our inten\u00adtions, the actions that arise from them and the con\u00adse\u00adquences that our actions have for us and oth\u00aders. This hand\u00adout explains what a deeper eth\u00adi\u00adcal exam\u00adi\u00adna\u00adtion can look like, which tools of applied ethics are avail\u00adable for this and ref\u00ader\u00adences to fur\u00adther read\u00ading: PDF file (276 kB)<\/a>, Docx file (75 kB)<\/a> <\/p>\n\nBibliography<\/h2>\n\n
Floridi, Luciano and Cowls, Josh.<\/strong> 2019. A Uni\u00adfied Fram\u00adwork of Five Prin\u00adci\u00adples for AI in Soci\u00adety. Har\u00adward Data Sci\u00adence Review. 2019, Vol. 1, 1, pp. 1\u201314. https:\/\/doi.org\/10.1162\/99608f92.8cd550d1<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"This also raises pro\u00adfound eth\u00adi\u00adcal ques\u00adtions: How can data pro\u00adtec\u00adtion and equal oppor\u00adtu\u00adni\u00adties be guar\u00adan\u00adteed? What respon\u00adsi\u00adbil\u00adity do uni\u00adver\u00adsi\u00adties bear in the use and devel\u00adop\u00adment of AI sys\u00adtems? And how can stu\u00addents and teach\u00aders be sen\u00adsi\u00adtized to a crit\u00adi\u00adcal approach to this tech\u00adnol\u00adogy? Deal\u00ading with eth\u00adi\u00adcal aspects is essen\u00adtial to ensure that AI is used [\u2026]<\/p>\n","protected":false},"author":1,"featured_media":7088,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"wp_typography_post_enhancements_disabled":false,"footnotes":""},"categories":[151],"tags":[],"class_list":["post-7100","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-hands-on"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/posts\/7100","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/comments?post=7100"}],"version-history":[{"count":0,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/posts\/7100\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/media\/7088"}],"wp:attachment":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/media?parent=7100"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/categories?post=7100"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/tags?post=7100"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}