I'm not on board with the smoking comparison generally, mostly because framing social concerns as public health issues has not successfully worked outside the regulation of cigarettes. The victory over smoking was a bit of a center-left fantasy -- the one "good" regulation that came out of an era of otherwise massive deregulation and privatization. In reality, we've decreased lung cancer rates somewhat while maintaining an overly harsh social stigma about addicts and smokers. Health care costs have not declined for anyone.

We forget all the other public health issues that have lost awareness since smoking regulation. Forever chemicals: only kindof regulated but not really! Air quality: Turns out the air gets bad and carcinogenic even when cigarettes aren't involved! Racism: declared a public health crisis in many places and now the backlash is so strong it's been wiped from federal science databases! COVID: Wtf happened there! Opioid addiction and addiction in general: still a massive public health issue that loses mindshare when we start talking about algorithmic social media in the same terms.

I find it curious that "reinforcement learning" is the specific facet of recommendation algorithms targeted here, especially in the context of public health. It would be incredibly difficult to regulate or excise. Reinforcement learning is present every corporate instance of knowledge management, marketing comms, or "business process optimization." It's in every video game built in the past 10 years in some form or other. It's everywhere; there's no way to specifically regulate the practice.

Additionally, it's becoming easier to develop our own alterna-platforms. Creating a content recommender system with custom attributes is no longer a pipe dream for smaller companies. Ideally, newer companies can demonstrate different outcomes for what a recommender system could be. (dreaming here, but it's possible)

For social media specifically, I'd encourage us to look more toward the models themselves as reflections of ideology. We could much more easily regulate ad targeting models built on lazy demographic assumptions borrowed from direct mail, or algorithms that prioritize individual engagement over content safety/relevance. We could look at algorithms that explicitly promote AI-generated engagement bait or any number of more specific factors that speak to how the technology is constructed to reinforce addictive digital behaviors.

I'm glad we're getting more finely tuned about approaches to regulation of social media, in the off chance that it happens in a way that doesn't exacerbate the fascism. But I think we need to let go of the cigarettes... because vapes and Zyns will find their way into the market... because smoking is the least of our worries... but mostly because in the U.S. we no longer agree on a collective commitment to support public health.