Lars Holmberg: Machine Teaching, a Human in the Loop Perspective on Machine Learning

Welcome to a K3 seminar with Lars Holmberg, Lecturer and PhD student in Interaction Design, K3/TS. The title of the seminar is:

Machine Teaching, a Human in the Loop Perspective on Machine Learning

It will take place on Wednesday, October 17 10.15-12.00 in The K3 Open Studio, NIC 0541, Niagara.

Here is an abstract for the talk.

Machine Teaching, a Human in the Loop Perspective on Machine Learning

Or: why you at the end of our time can’t say: I only delivered the milk.

On the second year of my PhD, a research area arises for me that combines many of my interests. In this talk I will paint a broad image of the landscape I have walked this far, and I hope to spark a discussion that helps me to move forward and focus my work for me, and/or at least equally important in the eyes of my supervisors :/.

Artificial Intelligence and Machine Learning opens up for a new type of products that often are described as disruptive to our way of living. Our prospect of handling this emerging technology is not altogether reassuring, this in the light of a 20th century that to a large degree relied on techno-fixes. Already the complexity of our connected society makes it increasingly difficult to anticipate the consequences of new technology; this in turn contributes to a technosocial opacity. This opacity indicates: the time when we could equal innovation with progress and handle the consequences afterwards is over.

Following my heritage from K3, I focus on use, users, design and technology during the whole design process. As a tentative approach in this field I have turned my attention to the relations we can build with Machine Learning agents. This shift is indicated by my use of the term “Machine Teaching” instead of “Machine Learning”. I see teaching as a relational activity that among other things builds on an understanding of the student’s prerequisites cultivated over time. This turns the focus from what ML-artefacts can learn towards a focus where we ask ourselves: What and how do we want to teach ML-artefacts if we see them as learners? This shift from a technology focus towards a human-technology relational focus might open a door to a technological development in this area through and for humanity.

I have planned my presentation as follows:

  • Tech presentation: a Machine Teaching prototype that uses state of the art tinker friendly Machine Learning technology (Fastai, Jupyter notebook).
  • Overview of the Artificial Intelligence and Machine Learning landscape and an outline over current research in in the area of Machine Teaching.
  • Areas surrounding a Machine Teaching thinking.

On the ethical and moral side, I will mainly present and discuss work by Shannon Vallor

On the human-technology relational side, I will mostly lean on postphenemonology.

  • Questions and discussion

Main references:

  1. Vallor, Technology and the Virtues: A Philosophical Guide to a Future Worth Wanting. 2016.
  2. Ihde, L. Langsdorf, K. Besmer, A. Hoel, and A. Carusi, “Postphenomenological investigations: Essays on human–technology relations,” 2015.
  3. Zhu, A. Singla, S. Zilles, and A. N. Rafferty, “An Overview of Machine Teaching,” 2018.

Leave a Reply

Your email address will not be published. Required fields are marked *