• Derek Lomas

Positive AI

Human civilization has lasted several thousand years. How might design theory help it last several thousand more? As artificial intelligence grows more powerful, we want to ensure a harmonious future.

Human values aren’t random — there are common factors that make us healthy and vital in a meaningful world. We need to ensure that powerful AI systems are constructed to align with our deep humanistic values.

When I came to Delft, I said that I had 2 intersecting goals:

  1. Designing smart systems that could use data to optimize outcomes and

  2. Understanding the right outcomes to optimize. In my thesis, I applied this to educational games — I showed how I could use AI algorithms to optimize time spent– but, how the optimization of time spent resulted in a game that had zero educational value. And thus, how we have to be careful about data-driven design.

Yet, assuming that AI systems and human systems will increasingly use metrics for system optimization (i.e., test scores, financial performance, number of likes, etc) there is a special need to develop valid metrics for what we actually value; we should “measure what we treasure.” When human values aren’t measured, they can be invisible to institutional systems. That’s why I care about improving the practical measurement of wellbeing in schools and in health care systems. Developing good measures can be seen as a mechanism for social change.

2 views0 comments

Recent Posts

See All