Stephen Downes

Knowledge, Learning, Community

Interesting look at the effect of data mining in education (8 page PDF). The author makes the point that research based in data mining works quite differently from traditional research. I quote:

  1. Data mining eschews the hypothetico-deductive process, relying instead on a strictly inductive process in which the model is developed a posteriori from the data itself.
  2. Data mining relies heavily on machine learning and artificial intelligence approaches, taking advantage of vastly increased computing power to use brute-force methods to evaluate possible solutions.
  3. Data mining characterizes specific cases, generating a predicted value or classification of each case without regard to the utility of the model for understanding the underlying structure of the data.
  4. Data mining aims strictly at identifying previously unseen data relationships rather than ascribing causality to variables in those relationships.

The author surveys the ethical implications of this. On the one hand, the good news is that model-based theories which treat all students as though they were the same are replaced with an approach recognizing the individuality of each student. But on the surface, the approach risks revealing information about students they don't want revealed, and risks fostering paternalism through the recommendation process, and at a deeper level, the risk of "scientism," or " he temptation to un-critically accept claims that purport to have scientific backing."

The current issue of the International Review of Information Ethics is a special issue on the digital future of education (it's issue number 21).

[Direct link]

Stephen Downes Stephen Downes, Casselman, Canada

Creative Commons License.

Copyright 2021
Last Updated: Mar 30, 2021 4:22 p.m.