Ten Heuristics for Mindful AI
Five years ago, I was supposed to be the keynote speaker for DesignUp a huge design conference usually held in India. I didn’t end up going to the conference — some visa snapfu — but I was interviewed by it’s conference magazine writer on responsible AI and ethics. See that interview here.
At that time I was distancing myself from the word “ethics,” because it felt to morality-laced. Also AI Ethics as a philosophy brought up complicated questions like whose ethics should be prioritized. Are principles ever universal and culturally aligned? And can there be one overall rule over AI that would take into account the customized, intersectional life we all live?
I didn’t think so. So I stuck with what I know — design and created what I then called Mindful AI principles that universally should be considered when designing with and for AI products, tools, platforms and experiences.
Here they are — The Ten Heuristics for Mindful AI
- Identify, document and root out trauma in data. Data is the love language of machines and its often where the ethical and irresponsible parts of AI begins and evolves. Data is infused with cultural, societal and racial trauma that riddled our communities and our individual lives. We create models with traumaticized data sets and those models have traumaticized outcomes for marginalized communities.
- Determine how data trauma and its shortcomings will affect model outcomes. This means playing out the…