In the public dis­cus­sion about Arti­fi­ci­al Intel­li­gence in Ger­man press texts, trust is often addres­sed lite­ral­ly, i.g.: EU wants to streng­then people’s trust in arti­fi­ci­al intel­li­gence (Darm­städ­ter Echo, 2021/04/27), Why Arti­fi­ci­al Intel­li­gence (AI) has a trust pro­blem (WELT Online, 2020/12/24) or Moreo­ver, gene­ral­ly available AI that does not reve­al its­elf could increase people’s dis­trust in tech­no­lo­gy (SPIEGEL Online, 2018/05/11) [Trans­la­ti­on: NK]. The­se head­lines and the respec­ti­ve artic­les seem to sug­gest that trust in AI appears to be the pre­re­qui­si­te for self-lear­ning sys­tems to be accept­ed by peo­p­le in a wide varie­ty of con­texts. But is this the case for every kind of self-lear­ning (i.e., AI) sys­tem?
Using a dis­cour­se-ana­ly­ti­cal approach, we exami­ne the ques­ti­on of which AI sys­tems in Ger­man public media dis­cour­se address trust and mis­trust. We iden­ti­fy sub-dis­cour­ses in which trust is overt­ly the­ma­ti­zed in con­nec­tion with AI and tho­se in which this is not the case. Does the the­ma­tiza­ti­on of trust, e.g., in rela­ti­on to machi­ne trans­la­ti­on play as big a role as in rela­ti­on to self-dri­ving cars? In which AI sub-dis­cour­ses is trust often addres­sed? In addi­ti­on, we are inte­res­ted in the ques­ti­on of whe­ther cer­tain key­words such as black box rein­force trust/mistrust in con­nec­tion with AI. The lexe­me black­box has nega­ti­ve con­no­ta­ti­ons in public dis­cour­se, while the term respon­si­ble AI in turn acqui­res posi­ti­ve con­no­ta­ti­ons through the posi­ti­ve-deon­tic adjec­ti­ve. In our con­tri­bu­ti­on to the panel, we thus aim to map the dis­cour­se on AI regar­ding its the­ma­tiza­ti­on of trust/mistrust and at the same time pur­sue the ques­ti­on of which key words are invol­ved in the lin­gu­i­stic con­sti­tu­ti­on of trust/mistrust.

Ein­ge­reicht von Dr. Moni­ka Hanaus­ka, Prof. Annet­te Leßm­öll­mann, Dr. Nina Kalwa