Introduction
When a system is made safer, its users may be willing to offset at least some of the safety improvement by using it more dangerously. A seminal example is that, according to Peltzman (1975), drivers largely compensated for improvements in car safety at the time by driving more dangerously. The phenomenon in general is therefore sometimes known as the “Peltzman Effect”, though it is more often known as “risk compensation”.[1] One domain in which risk compensation has been studied relatively carefully is NASCAR (Sobel and Nesbit, 2007; Pope and Tollison, 2010), where, apparently, the evidence for a large compensation effect is especially strong.[2]
In principle, more dangerous usage can partially, fully, or more than fully offset the extent to which the system has been made safer holding usage fixed. Making a system safer thus has an ambiguous effect on the probability of an accident, after its users change their behavior.
There’s no reason why risk compensation shouldn’t apply in the existential risk domain, and we arguably have examples in which it has. For example, reinforcement learning from human feedback (RLHF) makes AI more reliable, all else equal; so it may be making some AI labs comfortable releasing more capable, and so maybe more dangerous, models than they would release otherwise.[3]
Yet risk compensation per se appears to have gotten relatively little formal, public attention in the existential risk community so far. There has been informal discussion of the issue: e.g. risk compensation in the AI risk domain is discussed by Guest et al. (2023), who call it “the dangerous valley problem”. There is also a cluster of papers and works in progress by Robert Trager, Allan Dafoe, Nick Emery-Xu, Mckay Jensen, and others, including these two and some not yet public but largely summarized here, exploring the issue formally in models with multiple competing firms. In a sense what they do goes well beyond this post, but as far as I’m aware none of t
How much should 80k's US policy career advice change in light of the Supreme Court striking down Chevron deference? My quick take: In principle this is more power for the judiciary, but they still lack the power to initiate policy changes. So my quick take is it probably means someone who was undecided or weakly in favor of executive branch should try to get a job in Congress now. Should be a premium on experts who can write detailed legislation well now that that responsibility can't be delegated to the agencies. Interested in hearing others' thoughts.
Yeah my assumption is the same skills are in still in demand in roughly the same proportions but you'll just be working in a different building.
The main impact is if you were planning on working on implementing some rule that now faces serious constitutional questions and might be struck down.
One issue I missed above: because Congressional action typically requires consensus far more than the executive branch which is under the control of a single party, bipartisanship is now more important.
Current or former Congressional staffers, help me out in the DMs:
I see members of Congress fairly often for short periods of time, and I feel like most of my time with them is wasted in chitchat or just like...whatever's in the news. I wanna start using the time more effectively to promote the best available policy interventions at a given time. Assuming I will spend 5 minutes with a Congressperson per week (could be my Rep, could be others), and probably only have like 10 min available to spend on figuring out what to talk to them about, how do I efficiently figure out what I should be pushing that week? Should I just be a broken record about some specific thing no one is doing right now (esp given that I have access to Members other than my own; maybe this is a useful coordinating function), or should I focus more on already pending legislation?
Bonus points if you can suggest how to figure out timely appropriations that I should ask state/local elected officials to lobby for; I get wayyyyy more facetime with them.
Did you get any responses?