Human-Centered Machine Learning

Den 13 maj 2020 anordnades Halland Tech Meet, ett digitalt evenemang som samlar företagsledare, forskare och andra aktörer i Halland med fokus på digitalisering, teknologi och innovation. I detta avsnitt samtalar Anita Sant'anna (Research on the Go, och gäst i podden Episod 10), Joel Sandén (Region Halland) och Pontus Wärnestål (Högskolan i Halmstad och InUse) under ledning av Anna Petersson (projektledare för AI.m vid High5).

How can we systematically work with ethics and risks with services based on machine learning and other AI techniques? In this final episode, Cristofer and Pontus is talking to Vaike Fors, Associate Professor with a focus on design ethnography. Vaike has a long experience from working with design of services in the autonomous vehicles space.

Autonomous vehicles present a wide range of interesting design problems. Sigma Dolins, researcher at RISE's Sustainable Transport Team presents interesting cases surrounding dynamic ride-pooling, the experience of being both inside and outside of autonomous vehicles.

Healthcare is undergoing big changes on many fronts. One of the drivers for change is the race for health data and the many applications of data-driven services in all parts of the healthcare ecosystem. Anita Sant'Anna – PhD and consultant with over 12 years of research experience in signal and data analysis for health care applications – talks about some issues related to data collection and what human-centered design can bring to the healthcare sector.

In this episode, Jeanette kicks off the applied part of the course. Our guest is Martin Cooney, Assistant Professor in Artificial Intelligence and Robotics. Jeanette and Martin heads into the fast-moving space of EdTech, where AI techniques can help – and possibly hinder – learning and education.

In this episode Pontus talks to Chris Noessel, author of the book Designing Agentive Technology, about what designers need to know in order to design services that act on users’ behalf. This episode was recorded at the annual design conference From Business To Buttons, which feature international prominent speakers in the overlap between AI and Service Design.

In this episode, we’ll change gear a little bit. We’re discussing the role of human-centered design and its relation to data-driven services. We'll talk about how to consider users' mental models and expectations in our designs, how we move from rule-based tools to probability-based user experiences with "agentive" technology. We'll also talk about how to convey AI capabilities to the user gradually as they start using the service, and how to manage take-back and handoff between the system and the user. Important factors to consider is the "human-like" qualities in an AI-powered service and how we can achieve transparency and be able to explain the different service experience outcomes in different situations. Pontus will talk about this (and other things!) with Gustav Martner, CEO at ICA Reklam and founder of Digital Reliance.

Generative Adversarial Networks – or GAN – is powerful newcomer in the AI realm. GANs can be used to generate new data, that can help train new models as well as test and validate AI-based systems. Cristofer’s and Stefan’s guest this time is Edvin Listo Zec from RISE, who will discuss opportunities and risks with GANs.

“Deep Learning” marks the beginning of a new AI era, starting around 2012. Deep Learning is part of the reason AI has expanded so much recently. In this episode, Cristofer and Stefan discuss the basics of Deep Learning with Anders Holst from RISE.

Building machine learning applications require a lot of practical work and experience. Off-the-shelf solutions need to be tweaked and tailored to fit your organization’s needs and goals. The practicalities of implementing Machine Learning techniques is the topic of this episode. Stefan and Cristofer continue their discussion with Anders Holst from RISE.