Back button
Back

AI should account for what humans want

May 1, 2022
Graphic designer creative sketch planning AI application process development prototype wireframe for web mobile phone . User experience concept.

What makes us human not robots is the collective social knowledge that people build together. People want to be valued for their intuition, want to discover causality not just see patterns or correlations, want to connect with one another and share insights. Humans need agency. Humans will work around robots in order to have it. Design that fits human psychology takes account of a human’s motivation to find workarounds and do what a machine can’t do.

Suboptimal building of jobs can also block potential productivity gains from AI. Theoretically, if AI tasks and non-AI tasks are bundled, this could prevent specialization and lock up productivity gains. This is because workers will always choose to do something that AI cannot. There’s an opportunity cost of foregone potential non-AI (human) labor if the capital cost of AI is zero and firms give workers jobs where AI tasks and non-AI tasks are preset. That is, if there is no incentive or opportunity for humans to be creative and figure out what the AI can’t do as well and act accordingly, then the AI is also unable to be used to its best economic advantage.

If the AI is only used to replace humans, systems will become more precise over time but won’t allow for knowledge to evolve outside the algorithmic box. Conversely, when AI is used to nudge and augment human performance, there is no break in the link between doing and understanding.

The future of work is successful integration of both humans and machines. Humans will take responsibility for shaping how to tackle a problem or an opportunity; what data are needed, how to create the data, defining success, deciding on the acceptability of failure, designing what to do with outliers and responding when things go wrong.

The combination of human and machine intelligence is powerful — machines can help alert, nudge or correct a human and enhance people’s performance. But we have to account for the fact that humans need to feel a sense of control over machine interactions or risk either inefficiency, workarounds or rebellion.