Promises and Perils of Artificial Intelligence and Machine Learning: Disruption, Adoption, Dehumanisation, Governance, Risk, and Compliance

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 4 of 4
  • Item
    Stakeholder-dependent views on biases of human- and machine-based judging systems
    ( 2021-01-05) Mazurova, Elena ; Penttinen, Esko ; Salovaara, Antti
    Motivated by recent controversy over biases associated with algorithmic decision-making, we embarked on studying various stakeholders’ perceptions related to potential biases in verdicts from human-based and algorithm-based judging. In an empirical study conducted in the domain of gymnastics judging, we found that, while our informants viewed both human- and AI-based judging systems as being subject to biases (of different types), they were quite welcoming of a shift from human-based judging to machine-based judging. Our findings show that the athletes trusted strongly in unknown, “magic” capabilities of AI, thought to be more objective and impartial. This, in turn, encouraged potential acceptance of new technology. While the gymnasts saw AI-based systems in a positive light, judges demonstrated less favorable perceptions overall and less acceptance of AI technology, ex¬pressing concern about possible challenges of AI.
  • Item
    Machine Learning Systems in Clinics – How Mature Is the Adoption Process in Medical Diagnostics?
    ( 2021-01-05) Pumplun, Luisa ; Fecho, Mariska ; Wahl-Islam, Nihal ; Buxmann, Peter
    In a world with a constantly growing and aging population, health is a precious asset. Presently, with machine learning (ML), a technological change is taking place that could provide high quality healthcare and especially, improve efficiency of medical diagnostics in clinics. However, ML needs to be deeply integrated in clinical routines which highly differs from the integration of previous health IT given the specific characteristics of ML. Since existing literature on the adoption of ML in medical diagnostics is scarce, we set up an explorative qualitative study based on a conceptual basis consisting of the technological-organizational-environmental framework (TOE) and the healthcare specific framework of non-adoption, abandonment, scale-up, spread, and sustainability (NASSS). By interviewing experts from clinics and their suppliers we were able to connect both frameworks and identify influencing factors specific to the adoption process of ML in medical diagnostics.
  • Item
    Acceptance of AI for delegating emotional intelligence: Results from an experiment
    ( 2021-01-05) Aysolmaz, Banu ; Leyer, Michael ; Iren, Deniz
    Detecting emotions of other humans is challenging for us humans. It is however important in many social contexts so that many individuals seek help in this regard. As technology is evolving, more and more AI-based options emerge that promise to detect human emotions and support decision making. We focus on the full delegation of detecting emotions to AI to contribute to our understanding how such AI is perceived and why it is accepted. For this, we conduct an online scenario-based experiment in which participants have the choice to delegate emotion detection to another human in one group and to an AI tool in the other group. Our results show that the delegation rates are higher for a human, but surprisingly high for AI. The results provide insights that should be considered when designing AI-based emotion-detection tools to build trustworthy and accepted designs.
  • Item