BD.20.008 – Humane Helpline 2.0 – Empathic Responsible AI
Anyone with an unfulfilled need for social support can contact helplines, such as 113 Zelfmoordpreventie, De Kindertelefoon, MIND Korrelatie, De Luisterlijn, PratenOnline, and Fier (with their Centrum tegen Kinderhandel en Mensenhandel). They offer a unique social safety net by non-judgemental listening. However, helplines struggle to deal with: 1) the complexity of some help requests, 2) the training and support of the counsellors, and 3) the fluctuating 24/7 high demand for their services. Current ICT is unable to support the helplines: help seekers need human contact with real empathic understanding. ICT technology that can show empathy and understand the help seeker would be a societal game-changer. It requires scientific breakthroughs in learning human-verified dialogue patterns from chat corpus data, and runtime modelling of users (affect, cognitions, beliefs) and their situation during a chat. Additionally, it requires breakthroughs in data-driven dialogue systems for listening with comprehension and empathy, and in understanding user acceptance of these AI systems. The breakthroughs make possible that an AI can: 1) understand a help seeker and offer suggestions to the counsellor during a difficult conversation, 2) simulate realistic help seekers in a role-play training for counsellors, and 3) provide help seekers with an empathic understanding AI when no human is available. Furthermore, the breakthroughs are fundamental to a valuable infrastructure for digital helpline-services, but also for other NWA routes that require analyses-techniques for big natural language datasets. The sensitive nature of helpline data and service means that safe, secure, and responsible handling of data is vital. This project delivers machine learning algorithms for highly sensitive data, runtime help-seeker models, communication protocols and templates for dialogue systems, and conceptual models for acceptance and safety of humane digital help-services.
Keywords
conversational agents, e-health, empathic artificial intelligence, Helplines, mental health, Societal resilience
Other organisations
CWI, de Kindertelefoon, de luisterlijn, Fier (Centrum tegen Kinderhandel en Mensenhandel (CKM)), Hogeschool van Rotterdam., Leiden Universitair Medisch Centrum (LUMC), Leiden University (LEI), MIND Korrelatie, PratenOnline, Stichting 113 Zelfmoordpreventie, TNO, Universiteit Tilburg (TiU), Universiteit Twente (UT), Universiteit van Amsterdam (UvA), Vrije Universiteit Amsterdam (VU)
Submitter
Organisation | Delft University of Technology |
Name | Dr. W.P. (Willem-Paul) Brinkman |
w.p.brinkman@tudelft.nl | |
Website | https://www.tudelft.nl/ewi/over-de-faculteit/afdelingen/intelligent-systems/interactive-intelligence/people/current-group-members/willem-paul-brinkman/ |