Attention on the brain

Paper · Source
Social Media and AI

Attention has a practical and specific online context. It has to do with managing the flood of data and advertisements vying for your time, which is finite. The metadata 'exhaust' (feeds, clicks, links, search terms, etc.) you create can become a useful tool to help filter and funnel information so that your time isn't wasted on irrelevant or marginal information you encounter online. It filters what enters your field of attention, based on implicit and explicit data. Attention is also about people and participation (tip to Mary Hodder)--who you influence, who influences you, who you choose to pay attention to or to avoid.

There is also the notion of 'intention economy,' which Doc Searls says is about people engaging online with a clear notion of what they want. "I want my intention to buy or find something to be serviced by the system, whatever it becomes. Hearing 50 companies tell me how they are competing for my attention doesn’t cut it; they are still looking at me as an eyeball,"  Doc said. It inverts the traditional advertising model.

Technorati CEO Dave Sifry extended that idea succinctly during the Search SIG discussion: "In the world of eyeballs, a good consumer is someone tied to chair consuming content and crapping cash. Participating in the growth of a site or community is about contributing to something larger than myself and receiving something from others."

Lines of inquiry this paper opens 17

Research framings built by reading the notes related to this paper — the questions it feeds into.

How can humans maintain meaningful oversight as AI systems become increasingly autonomous and complex? How does AI-generated content undermine authentic engagement on social platforms? What factors drive AI persuasiveness and how can it be mitigated? How do false presuppositions and sycophancy drive persistent false beliefs in models? Does transformer attention architecture inherently drive sycophancy? Why do some clarifying approaches produce understanding while others just satisfy? Do language models lack essential therapeutic presence and engagement? How do social dynamics distort aggregated online ratings? How do spurious versus genuine rewards shape model reasoning and behavior? What determines appropriate intervention timing and manner for AI agents?