The Volokh Conspiracy

You have job interviews with two employers and are turned down in both of them. At the next one you are asked if you have had any prior interviews. You recount your unhappy recent history and the employer concludes that the two prior rejections probably meant something. This helps him decide to pass on you. The process continues and accelerates from there; the next interviewer finds you have three rejections and has even more cause for concern than the previous one.

This is an example of an information cascade. Notice that all the interviewers might be acting rationally. If you feel uncertain about something, it might make sense to defer to others who have already decided; maybe they knew more than you do. And if the next player likewise has no firm basis for decision it might be entirely reasonable for him to see the growing agreement, find it impressive, and go along. But whether reasonable or not, the result is that the belief gains a kind of empty momentum: there is growth in its acceptance but not in its likelihood of being true, which hasn't changed and may be small.

There are a number of implications for law. One involves the hazards of sequential decisionmaking. When witnesses are asked what they saw, they say different things alone than if they first hear how others describe the events; there are various ways to interpret this, one of which is that it's a kind of cascade. That is why the federal guidelines say that witnesses to crimes should be separated and shouldn't talk to each other.

A similar problem arises when jurors vote on a case. Should they vote simultaneously or sequentially? A simultaneous vote has the advantage of avoiding cascades: we don't want the third juror swayed by what the first two say, then the fourth juror swayed by what the first three say, and so on. So the choice of procedure might matter; oddly, though, we leave it up to each jury to decide what to do. The same general question arises again when judges vote. On the United States Supreme Court the Justices vote openly and one at a time, starting with the Chief Justice and then descending to the most junior member, who already knows how all the other Justices voted when his turn arrives -- suggesting a danger of cascades.

The risk of cascades repeats on a larger scale in elections. Think about cascades created by early primaries, for example, or by public opinion polls (which is why some countries ban them in the days or weeks before an election). Might a cascade also arise when courts in different jurisdictions are presented with the same question, one after the other? There are other interesting kinds of cascades, too, but this entry is getting long; so if you'd like to read about them (or more about the kind just described), they are the subject of one of the sample chapters available at the web site.

This post wraps up my guest-blogging for the week. I hope these discussions have piqued the interest of some of you in The Legal Analyst. The sample chapter just mentioned, and the others at the web site, should give you a clearer idea of whether the book is for you (they go into more detail than I've had space to do here); and there will soon be a fourth sample chapter available at Eugene's own site — the chapter on slippery slopes, which, as I mentioned before, I co-wrote with Eugene but is really just a short adaptation of his classic article on the subject. Thanks so much to Eugene and the other writers at his site for having me here, and to the many readers who have sent me suggestions or posted them to the comment threads.

	[Ward Farnsworth (guest-blogging), August 3, 2007 at 10:44am] Trackbacks Cascades. You have job interviews with two employers and are turned down in both of them. At the next one you are asked if you have had any prior interviews. You recount your unhappy recent history and the employer concludes that the two prior rejections probably meant something. This helps him decide to pass on you. The process continues and accelerates from there; the next interviewer finds you have three rejections and has even more cause for concern than the previous one. This is an example of an information cascade. Notice that all the interviewers might be acting rationally. If you feel uncertain about something, it might make sense to defer to others who have already decided; maybe they knew more than you do. And if the next player likewise has no firm basis for decision it might be entirely reasonable for him to see the growing agreement, find it impressive, and go along. But whether reasonable or not, the result is that the belief gains a kind of empty momentum: there is growth in its acceptance but not in its likelihood of being true, which hasn't changed and may be small. There are a number of implications for law. One involves the hazards of sequential decisionmaking. When witnesses are asked what they saw, they say different things alone than if they first hear how others describe the events; there are various ways to interpret this, one of which is that it's a kind of cascade. That is why the federal guidelines say that witnesses to crimes should be separated and shouldn't talk to each other. A similar problem arises when jurors vote on a case. Should they vote simultaneously or sequentially? A simultaneous vote has the advantage of avoiding cascades: we don't want the third juror swayed by what the first two say, then the fourth juror swayed by what the first three say, and so on. So the choice of procedure might matter; oddly, though, we leave it up to each jury to decide what to do. The same general question arises again when judges vote. On the United States Supreme Court the Justices vote openly and one at a time, starting with the Chief Justice and then descending to the most junior member, who already knows how all the other Justices voted when his turn arrives -- suggesting a danger of cascades. The risk of cascades repeats on a larger scale in elections. Think about cascades created by early primaries, for example, or by public opinion polls (which is why some countries ban them in the days or weeks before an election). Might a cascade also arise when courts in different jurisdictions are presented with the same question, one after the other? There are other interesting kinds of cascades, too, but this entry is getting long; so if you'd like to read about them (or more about the kind just described), they are the subject of one of the sample chapters available at the web site. This post wraps up my guest-blogging for the week. I hope these discussions have piqued the interest of some of you in The Legal Analyst. The sample chapter just mentioned, and the others at the web site, should give you a clearer idea of whether the book is for you (they go into more detail than I've had space to do here); and there will soon be a fourth sample chapter available at Eugene's own site — the chapter on slippery slopes, which, as I mentioned before, I co-wrote with Eugene but is really just a short adaptation of his classic article on the subject. Thanks so much to Eugene and the other writers at his site for having me here, and to the many readers who have sent me suggestions or posted them to the comment threads. Related Posts (on one page): Slippery Slopes in 10 Pages: Cascades. Puzzles About Proof. The Stag Hunt. Don't Shoot Me, I'm Only the Piano Player. Prof. Ward Farnsworth's The Legal Analyst: