Navigating the Ethical Crossroads: Bridging the gap between Predictive Power and Explanation in the use of Artificial Intelligence in Medicine

Read the Paper

Reference:

Riva, G., Sajno, E., De Gaspari, S., Pupillo, C., & Wiederhold, B. K. (2023). Navigating the Ethical Crossroads: Bridging the gap between Predictive Power and Explanation in the use of Artificial Intelligence in Medicine. Annual Review of CyberTherapy and Telemedicine, 21(A), 3-7.

Artificial Intelligence (AI) has emerged as a transformative force in medicine, provoking both awe and apprehension in clinicians and patients. In fact, the challenges posed by medical AI extend beyond mere technological hiccups; they delve into the very core of ethics and human decision-making. This paper delves into the intricate dichotomy between the clinical predictive prowess of AI and the human ability to explain decisions, highlighting the ethical challenges arising from this disparity. While humans can elucidate their choices, AI often operates in opaque realms, generating predictions without transparent reasoning. The paper explores the cognitive underpinnings of prediction and explanation, emphasizing the essential interplay between these processes in human intelligence. It critically analyzes the limitations of current medical AI systems, emphasizing their vulnerability to errors and lack of transparency, especially in a critical domain like healthcare. In this paper, we contend that explainability serves as a vital tool to ensure that patients remain at the core of healthcare. It empowers patients and clinicians to make informed, autonomous decisions regarding their health. Explainable Artificial Intelligence (XAI) tackles these challenges. However, achieving it is not easy, and it is strongly dependent from different technical, social and psychological variables. Achieving this objective highlights the urgent requirement for a multidisciplinary approach in XAI that integrates technological knowledge with psychological, cognitive and social perspectives. This alignment will foster innovation, empathy, and responsible implementation, shaping a healthcare landscape that prioritizes both technological advancement and ethical considerations.

Detection of Break in Presence in Full Body Illusion using Machine Learning

Read the Paper

Reference:

De Gaspari, S., Sajno, E., Di Lernia, D., Brizzi, G., Sansoni, M., & Riva, G. (2023). Detection of Break in Presence in Full Body Illusion using Machine Learning. ANNUAL REVIEW OF CYBERTHERAPY AND TELEMEDICINE 2023, 28.

Break in presence (BIP) is a concept that refers to that condition in which the person immersed in a virtual reality (VR) environment disengages from it, due to various factors, and focuses more on the real environment. BIP occurs due to certain factors that may be related to errors in the sensory data of the virtual reality environment (VRE). These errors can influence the multisensory integration process by generating a multisensory conflict, which results in BIP phenomena. Several studies have identified a correlation between BIP and the physiological (e.g., electrocardiogram/ ECG) responses of people, showing that there is a correlation between these two aspects. In this study, we used the physiological (ECG) data collected in a previous study, in which participants were involved in a Full Body illusion (FBI) experiment, i.e. a paradigm that uses VR to generate a synchronous multisensory stimulation condition and an asynchronous multisensory stimulation condition. The FBI asynchronous stimulation condition can generate a multisensory conflict that is the same phenomenon underlying BIP. In this study, we used these physiological (ECG) data, collected throughout the FBI, to train different machine learning (ML) models to be able to detect the multisensory conflict underlying BIP from ECG tracings alone. The results showed that two ML models, Decision Tree and Random Forest, are able to detect the multisensory conflicts underlying BIP.

Who can help me? Reconstructing users' psychological journeys in depression-related social media interactions

Read the Paper

Reference:

Morini, V., Citraro, S., Sajno, E., Sansoni, M., Riva, G., Stella, M., & Rossetti, G. (2023). Who can help me? Reconstructing users' psychological journeys in depression-related social media interactions. arXiv preprint arXiv:2311.17684.

Social media are increasingly being used as self-help boards, where individuals can disclose personal experiences and feelings and look for support from peers or experts. Here we investigate several popular mental health-related Reddit boards about depression while proposing a novel psycho-social framework. We reconstruct users' psychological/linguistic profiles together with their social interactions. We cover a total of 303,016 users, engaging in 378,483 posts and 1,475,044 comments from 01/05/2018 to 01/05/2020. After identifying a network of users' interactions, e.g., who replied to whom, we open an unprecedented window over psycholinguistic, cognitive, and affective digital traces with relevance for mental health research. Through user-generated content, we identify four categories or archetypes of users in agreement with the Patient Health Engagement model: the emotionally turbulent/under blackout, the aroused, the adherent-yet-conflicted, and the eudaimonically hopeful. Analyzing users' transitions over time through conditional Markov processes, we show how these four archetypes are not consecutive stages. We do not find a linear progression or sequential patient journey, where users evolve from struggling to serenity through feelings of conflict. Instead, we find online users to follow spirals towards both negative and positive archetypal stages. Through psychological/linguistic and social network modelling, we can provide compelling quantitative pieces of evidence on how such a complex path unfolds through positive, negative, and conflicting online contexts. Our approach opens the way to data-informed understandings of psychological coping with mental health issues through social media.