adesso study on trustworthy AI: Companies would like more control over AI

Whether it’s searching for new materials, identifying fraudulent invoices or driving autonomous vehicles, the tremendous potential of AI is already being demonstrated by thousands of applications across all industries. Certain decisions based on intelligent algorithms have a direct effect on people’s lives – such as when a recommendation by a software program influences the selection of personnel or indicates that a bank should reject a loan application. In these cases, it is up to those responsible to shed light on the darkness surrounding AI decision-making. This means identifying the origin and characteristics of the training data used by an AI system, assessing how transparent and understandable the algorithm is and reviewing compliance with regulatory rules and ethical principles. The EU Artificial Intelligence Act, which was adopted in April 2021, also calls for transparency in the field of AI. It says that users have a right to be informed about the type of AI they are interacting with.

adesso asked the following questions on the subject of trustworthy AI to 500 decision-makers who work for companies using AI applications and are involved in system development: how important is trustworthy AI? What do you think users’ attitudes are towards AI applications? What kind of regulation is necessary in your view?

The main findings of the study are outlined below.

• Considerable trust in AI. Most respondents have high expectations for AI applications, with 88 % believing that AI will help them make better decisions. They are equally optimistic about users’ attitudes to the technology. 61 % are convinced that consumers regard AI as fundamentally trustworthy.
• Regulation needed. Nevertheless, 80 % of those surveyed think that AI should not be used without human control. In addition, 54 % are of the opinion that the use of AI technologies should be more strictly regulated. Such calls are backed by as many as 75 % of respondents from the banking and financial services sectors. A total of 70 % of all participants are in favour of setting up an internal AI ethics committee. There is strong support for the idea of a dedicated expert panel in the healthcare sector (87 %) and in public administration (83 %).
• AI in the workplace. The respondents believe that companies must involve their staff in AI projects to a greater extent, as this is the only way for them to develop the necessary trust in the technology. Of those surveyed, 70 % think that preparing their employees to collaborate with AI applications will be a major challenge. Participants from the automotive sector see the situation as particularly problematic: 91 % fear that it will be difficult to convince their teams to embark on human-AI cooperation.

adesso and the Cologne-based market research agency heute & morgen conducted a survey on the subject of trustworthy AI in May 2022. The group of participants comprised 500 decision-makers in Germany who work for companies with AI applications.

The study can be downloaded for free: https://ki.adesso.de/de/topics/trustworthy-ai.html