AI gets the benefit of the doubt, but it needs control mechanisms.
How can those in charge make decisions made by an AI application transparent? How do they manage potential risks of AI use? The answers to these questions are essential for consumer acceptance of the technology. As part of a representative study, the IT service provider adesso surveyed AI experts in companies on the topic of “Trustworthy AI“. According to the survey, the majority believe that AI is trusted by customers. At the same time, the participants in the study would like to see more control mechanisms.
Shedding light on AI decisions
Some of the decisions based on intelligent algorithms affect people’s lives in a direct way. For example, when an assistance software’s recommendation affects personnel selection. Or ensures that a bank rejects a loan application. In these cases, those responsible must shed light on the AI decisions. This ranges from the origin and characteristics of the training data used by an AI, to the transparency and comprehensibility of the algorithm, to compliance with regulatory requirements and basic ethical values. AI transparency is also required by the EU Artificial Intelligence Act of April 2021, which states that users have a right to be informed about what kind of AI they are interacting with.
Adesso surveyed 500 decision-makers who work with AI applications in companies and develop the systems on the topic of Trustworthy AI. The study’s key findings:
High trust in AI
The majority have high expectations for the application. For example, 88 percent hope to be able to make better decisions through the use of AI. The participants in the study are just as optimistic when it comes to the question of how users feel about the technology. 61 percent are convinced that consumers consider AI to be fundamentally trustworthy.
Nevertheless, four out of five of those surveyed think AI should not be used without human control as a matter of principle. 54% also believe that the use of AI technologies should be more closely regulated. In the banking and financial services sector, as many as three quarters support this claim. The idea of a separate council of experts is particularly popular in healthcare (87%) and public administration (83%).
AI in the workplace
From the perspective of the study participants, companies need to involve employees more in AI projects. This is the only way they will develop the necessary trust in the technology. Seventy percent of respondents see a major challenge in preparing their own workforce to work with AI applications. Study participants from the automotive industry are particularly critical in their assessment of the situation: In this case, 91% fear that it will be difficult to convince their own teams of the benefits of human-machine collaboration.
Adesso conducted the survey together with the Cologne-based market research agency heute & morgen in May 2022. The study is available free of charge via download: https://ki.adesso.de/de/topics/trustworthy-ai.html