T6-EX1-4 - Machine learning introduces new perspectives to data agency in K–12 computing education

3. Research Full Paper
Matti Tedre1 , Henriikka Vartiainen1, Juho Kahila1, Tapani Toivonen1, Ilkka Jormanainen1, Teemu Valtonen1
1 University of Eastern Finland

The directions that computing and information technology took in the 2000s have brought the concept of data literacy and its many variants into limelight. Pervasive computing, data-intensive analysis, cloud computing, social media, and the Internet of things have enabled tracking and profiling of people at massive scale both in the physical and virtual realms. Users of apps, gadgets, and web services leave traces that either directly reveal, or can be used to infer, the users’ media preferences, moods, political affiliations, and many of their secrets.  This has created pressure towards renewing computing education at all levels.

The popular term “data literacy” involves, for instance, understanding what data one creates, what happens to them, and with what consequences. A more active concept, data agency, refers to people's volition and capacity for informed actions that make a difference in their digital world. Data agency extends the concept of data literacy by emphasizing people’s ability to not only understand data, but also to actively control and manipulate information flows and to use them wisely and ethically. What citizenship in the 2000s requires is not just passive knowledge of information flows that surround people and influence their behavior, but active ability to take control of those flows and harness them for use. The datafied world requires also understanding of the many facets of automation, including data-driven machine learning and rule-based programming.

This article describes the theoretical underpinnings of the "data agency" concept for K–12 computing education. It discusses the epistemological and methodological changes driven by data-intensive analysis and machine learning. Epistemologically the many new modalities of automation non-reductionist, non-deterministic, and statistical; the models they rely on are soft and brittle. This article also presents results and new perspectives from three pilot studies on how to teach central machine learning concepts and workflows in the primary school and secondary school. Data were collected in three different schools over three series of workshops, in which children ideated and designed machine learning-based apps, trained the machine learning models, and tested the final products. The results show marked change in children's data agency, as well as in their understanding of principles and workflows of machine learning. The article presents lessons learned from participatory making and learning machine learning concepts through co-creation of tensor flow-driven solutions.