Beyond Principles: A Process for Responsible Tech
Meet the new design process: An upgrade with ethics & digital wellbeing baked in.
In the frenzied quest to keep a moral handle on AI, dozens of groups have organized the collaborative creation of ethical principles. From the Asilomar AI Principles to the IEEEs Ethically-aligned design specification, to the individual efforts of companies like Google and Microsoft to set their philosophical leanings onto paper. We are awash in principles.
While principles are essential, many tech makers are frustrated by how little help they provide in actual practice. Principles must be sufficiently abstract to retain truth across contexts, but this abstraction also leaves them too vague to be useful for specific design decisions on their own.
For example, Microsoft sensibly states “AI systems should treat all people fairly” but what is fair? Is affirmative action fair? Is it fair to violate the rights of an individual on behalf of the many? Anand Rao and Ilana Golbin of PwC point to the fact that “There are at least 20 mathematical definitions of fairness, and when we choose one, we violate some aspect of the others. In other words, it is impossible for every decision to be fair to all parties.” So how exactly might we go about using a principle like this to make a design decision?
Beyond ethical principles, tech makers need actionable methods that fit into their real world practice.
Of course, design contexts are so unique and various that no simple list of guidelines could ever be useful across projects. In other words, you can’t give people ethical design rules to apply everywhere. Design decisions need to be made with sensitivity to context, goals, culture, vision and values.
What you can provide is process.
Responsibility via process
A handful of pioneering efforts to develop tools and collaborative practices for more responsible design have made important headway (see doteveryone, ethicskit.org, IEEEs specifications and Tristan Harris’ Centre for Humane Design). But how do we fit these into our practice? How much do we need to rehaul the way we do things, and how do we know it will work?