congrats!!!
Can the legal system adapt to generative AI? There are clues in unexpected places!
I'll be revealing those clues, and what they tell us about the future of law & AI, at the Hanley Democracy Center @univmiami.bsky.social Monday at 6:30 PM.
Details: events.miami.edu/event/artifi...
See you there!
LLMs are going to prove that we can quantify legal work.
But Goodhart's law will make us wish we couldn't.
am i right to assume that you're a celtics fan delighted to see kobe's 81 fall down the list?
lololololol
our vice dean just started a monthly essential updates newsletter and it's called "constructive notice" and if you are not a legal academic i can assure you that this is one of the funniest things you can possibly imagine
i realize that i am the target market of "people who lived in nyc during peak lcd soundsystem."
still: i am very much enjoying kiss all the time.
Looks great! Will read with interest.
What’s the article? Send me a draft!
To close, there’s a lot of opportunity to think about how to align private incentive with social welfare when research entails discovery of many dead-ends. All of our proposals have flaws. Details in the paper—get it here!!
buff.ly/XUnO9Eh
3) Also, prizes typically consist of competitors racing to be first to uncover positive info. We can add “red teams” to the competition—these participants would race to be the first to demonstrate that a possible approach won’t work.
2) Prizes are typically structured competitions with internal governance, not simply one-off rewards at the end of a race. We can leverage this by mandating disclosure of negative info among participants.
Finally, we think prizes are most promising, & offer suggestions for how to improve them.
1) Prizes for success are well-known. We should offer prizes for failures too! At least, we should when failure is verifiable. Vulnerability rewards programs (i.e., “bug bounties”) offer proof-of-concept.
Even with intermediaries, tho, situation is suboptimal bc info does not flow readily & firms cannot be fully compensated for social value. Still, this adds to explanations for why innovation ecosystems are geographically clustered, industry-specific, & hard to displace.
In tight-knit innovation ecosystems, intermediaries use reputation & trust to facilitate transactions over negative info. Think Sherwood Partners in SV, acqui-hires, etc.
What to do?
Property solutions are not great bc they limit disclosure. So we expect lots of duplicative research when trade secrets & similar tools can “protect” negative info.
We show how these effects interact under diff policies. For benchmark (no protection for negative info):
1) Firms delay investing bc they expect to learn negative info from investments by others.
2) Firms underinvest bc they can’t capture value of negative info to later firms.
Focus on opposing effects:
1) Info benefit effect --> discourages entry. If earlier firm fails, later firms have better odds of success bc they can avoid dead-end.
2) Lost opportunity effect --> encourages entry. Waiting firms could lose chance to participate if earlier one succeeds.
So Firm A’s investments won’t reflect social value bc it can’t capture the value that failure confers on its rivals.
But! All firms realize they can take advantage of prior discoveries of dead-ends. So they all want to wait before investing.
We formalize intuitions abt this problem.
The problem: info about failure is a public good, just like info about success.
Suppose we want to find new battery materials. If Firm A discovers that one material won’t work, its rivals would all benefit from that knowledge—at min, they can avoid the dead-end.
We explore the well-known but undertheorized value of failure in research. Key insights:
1) Policy must account for timing, not just level, of investment in risky research.
2) Solutions must encourage disclosure of failure. As a result, subsidies > property.
New paper 📣🎉🧵
Thrilled to share that my next article, Rewarding Failure, will be published in J Law Econ & Org!!
Was absolutely delightful to work with the brilliant Francesco Parisi & Alice Guerra on this one.
buff.ly/nqtoDfQ
"weird"
This is one of my favorite secret facts about the world
📢 Equitable Growth was recently awarded a grant as part of a new MacArthur Foundation initiative!
The award provides momentum for EG's AI program, which invests in research and policy analysis to address questions at the intersection of AI and the economy.
More 🔗:
Join us at the Virtual #AI Lecture Series on 2/20, featuring @apublicgood.bsky.social, director of #MiamiLaw's Business of Innovation, Law & Technology concentration! This ongoing series dives into the real-world impact of AI across research, business, and daily life. RSVP today!
And here I thought I was the last person still listening to details of the war at max possible volume
One of our top requests lately is to develop an MCP server. This will allow you to seamlessly access our APIs from tools like ChatGPT and Claude. We're working on it and eager to hear your input: www.courtlistener.com/help/mcp/
One of my students just tested positive for Claude Code.
Be careful out there folks . . . this is a highly contagious condition. Wear a mask and practice social distancing. If we work together, we can flatten the curve.
whale song ayfkm 🇺🇸🇺🇸🇺🇸
