But when pressed on a specific example, fewer were up for it, and the demographics flipped. Men, who had been relatively anti-breach in general, were pro-breach in the particular case. Something about the general nature of eliciting questions about promissory behavior was tripping everyone's wires.

Here, we asked a general Q: is breach good, sometimes?, and then a more specific one, would I breach a particular agreement?

Surprisingly, to us, most people reported saying that sometimes breach is good. They didn't look particularly formalist.

The big idea is that the field's orientation is built on the litigated cases, which hold up an increasingly distorted mirror to commercial life. But -- contrary to the prevailing account -- that doesn't mean we need (or can) refashion doctrine to return us to the contracting practices of the 1800s.

But real move of the paper is to connect this categorizing project with attitudes. Our thesis is that you learn from the deals you are in (which aren't randomly distributed!)

The baseline here is that for many deal types, Americans are anti-breach formalists.

Other findings were less obvious: parents report doing more bigger-stakes contracting of all kinds.

And black respondents were disproportionately likely to engage in gig economy contracting.

But this isn't always true. Sometimes, as we show using a survey, being in some kinds of contracts (gig work) makes you more up for ignoring their explicit terms. That is, sometimes experience is a pragmatic teacher.

We divide contracting life into three rough types: the old world (think, the opinions in the casebooks), the real world (aversive consumer contracts), and the new world (gamified gig contracting).

One thing we see is that the new world contracts look *real* different than other types.

For instance, here's two of dozens of charts and tables, offering a comparative sense of how common certain kinds of deals are, and how some of those experiences vary by race, gender and self-reported class.

Knew it.

I think we're getting somewhere.

A good article on this:

columbialawreview.org/wp-content/u...

And the other, unjust enrichment claims regarding a surprisingly successful flashlight company.

No, Luke. The shadow language came from the festival organizers themselves. palestinewrites.org/2023/09/15/l...

One super cool thing we can do is show how pieces of evidence contribute to the model's confidence. Thus, in Stewart v. Newbury, the model reaches an outcome built on a contract and two different extrinsic pieces of meaning, one materially, and the other marginally, important.

And here's a way to get under the hood of a model: freezing it mid-sentence as it predicts the meaning of a pre-nup agreement in Famiglio v. Famiglio.

We explain the use, and deep the case, for LLMs in law: they can offer subtle predictions about the meaning of words in context. It's not just a chatbot. Here are multiple models describing the best reading of famous language from the 9th circuit's Trident case.