WP 3 NACHOS | Area: Implementation
Key question of the sub-project:
How can acceptance be increased and resistance reduced when introducing innovations?
WP 3: Reducing resistance and increasing acceptance of innovative,
intelligent systems
(PI: Jun.-Prof. Dr Kai Heinrich | Co-PIs: Prof. Dr Heike Ohlbrecht, Prof. Dr Elmar Lukas)
In contrast to traditional IT systems as part of information systems, intelligent systems – as innovations featuring fully or partially autonomous decision-making logic – represent, on the one hand, an innovative asset for all sectors of the economy and society; on the other hand, due to their complexity and opacity, they create a barrier between humans and the system. In most application scenarios, successful collaboration between humans and intelligent systems depends on the design of a hybrid, socio-technical information system that can combine the best of both worlds. However, this can only be achieved if the people involved in the decision-making process accept the intelligent system. Whilst, for example, artificial intelligence-based decision support systems in the medical sector deliver impressive results in the prediction and detection of diseases, the integration of these results into the clinical decision-making process is often hampered by an aversion to algorithmically generated predictions on the part of human stakeholders – in this case, medical professionals and, not least, the patients themselves.
Whilst the reasons for this aversion can be very varied, research into them has largely been conducted at the human level, which does not allow for the direct derivation of design guidelines for hybrid, intelligent systems or for team-building between humans and systems. To this end, theoretical findings on the topic of algorithmic aversion must first be operationalised and then empirically tested.
Within the scope of this sub-project, there are three specific areas to be examined: (a) firstly, the involvement of humans as decision-makers in the development (specifically the training) of these systems may help to mitigate the loss of control associated with algorithmic aversion. To this end, various interactive approaches to system training will be enhanced with ‘human-in-the-loop’ elements and tested. Furthermore, (b) the lack of transparency regarding the intelligent system’s decision-making logic will be addressed, and various methods for explaining this logic to human stakeholders will be examined. Finally, (c) the implementation process of such systems is examined as users’ first point of contact with this innovation, and various variants of a ‘familiarisation phase’ are tested.
In empirical experiments across various fields of application (including industry and medicine), the effects of the measures outlined in (a)–(c) on people’s aversion to and acceptance of intelligent systems will be tested.