{"items": [{"author": "Elizabeth", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673469040522", "anchor": "fb-673469040522", "service": "fb", "text": "In addition to finding your calculations interesting, I took the test, and they believe I am incredibly likely to side with the defendant. However, not knowing the facts of the case, from the description I am in fact extremely biased toward the plaintiff. So I'm trying to think about whether I'm an outlier, or whether there's something that seems flawed about the questions.", "timestamp": "1408712261"}, {"author": "Danner", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673469359882", "anchor": "fb-673469359882", "service": "fb", "text": "Jeff, each lawyer might be acting with imperfect information while deciding if a jury member is on their side. It's like when republicans get low income people to vote for them against their own best interests.", "timestamp": "1408712584"}, {"author": "Will", "source_link": "https://plus.google.com/115623320734956550423", "anchor": "gp-1408713437625", "service": "gp", "text": "The opposite of \"more likely to side with plaintiff\" isn't \"more likely to side with defendant,\" it's \"not more likely to side with plaintiff.\" These bars represent two things: 1. How much of an impact does the evidence have? 2. In which direction does that impact move the expected outcome? Income over 50K simply has no impact while income under 50K does. This is not a problem.", "timestamp": 1408713437}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673470337922", "anchor": "fb-673470337922", "service": "fb", "text": "@Danner: Imperfect information can't account for this. If learning X makes you expect more Y than you would have, learning not-X must make you expect less.", "timestamp": "1408713453"}, {"author": "Danner", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673471131332", "anchor": "fb-673471131332", "service": "fb", "text": "Think about heads up poker. One guy has a full house, the other has a flush. They each have such a rare hand that they are each more certain that they won. It's like they are operating on two scales, one dealing with likelihood that you are on their side, and the other dealing with the accuracy of the assessment. Yes, we know that that last spade gave the full house to the opponent, but overall it still looks like the flush in spades is good.", "timestamp": "1408714171"}, {"author": "Jeff Kaufman", "source_link": "https://plus.google.com/103013777355236494008", "anchor": "gp-1408714497458", "service": "gp", "text": "@Will\n\u00a0\"The opposite of 'more likely to side with plaintiff' isn't 'more likely to side with defendant,' it's 'not more likely to side with plaintiff.'\"\n
\n
\nIf you're on a jury, and you will eventually be asked to make a decision, you will have to side with either the the defendant or the plaintiff. \u00a0The opposite of \"more likely to side with the plaintiff\" is \"less likely to side with the plaintiff\" and that's equivalent to \"more likely to side with the defendant\".\n
\n
\n\"Income over 50K simply has no impact while income under 50K does.\"\n
\n
\nThis is just not possible. \u00a0Imagine I'm going to flip a coin. \u00a0You see me put two coins into an empty bag. \u00a0One coin has heads on both sides, one is an ordinary fair coin. \u00a0I draw one coin out of the bag at random, flip it, and cover it so neither of us can see it. \u00a0How likely is it to be heads? \u00a0Well, either it's the heads-heads coin in which case 100% or it's the heads-tails coin in which case 50%, so we would both guess it's 75% likely to be heads.\n
\n
\nThen you look in the bag, and you learn which coin we didn't use. \u00a0If the one in the bag is heads-heads then the one I flipped was heads-tails and your estimate goes from 75% to 50%. \u00a0Otherwise your estimate goes from 75% to 100%. \u00a0Getting one answer would make you think \"more likely to be heads\" while getting the other would be \"less likely to be heads\".\n
\n
\nYour claim instead is that \"learning that the coin you flipped is a fair coin simply has no impact on your prediction while learning that the coin was heads on both sides does\", which is clearly not the case.", "timestamp": 1408714497}, {"author": "Tobias", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673471685222", "anchor": "fb-673471685222", "service": "fb", "text": "If household incomes under 50k would be a tiny minority, it would be possible for having over 50k only having a tiny effect. The more people you have the less weight each of them has to carry.", "timestamp": "1408714562"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673473331922", "anchor": "fb-673473331922", "service": "fb", "text": "True, Tobias, but alas I don't think they're a tiny minority.", "timestamp": "1408715821"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673474130322", "anchor": "fb-673474130322", "service": "fb", "text": "Jeff, regardless of the text, it looks like maybe every question after the first has at least one choice that doesn't move the slider at all. So in that sense almost every question has the problem you're pointing to.", "timestamp": "1408716216"}, {"author": "Matt", "source_link": "https://plus.google.com/113951350991359002027", "anchor": "gp-1408716446429", "service": "gp", "text": "Of course this is possible.\n
\n
\nImagine the defense lawyer's opinion-vs-income graph is linear, from a high at $0 then zero at some high value (let's say $100k), and the prosecution's graph has 3 segments: flat from $0 to $50k, then increasing to $100k, then flat again. Further let's assume that the 2 diagonal lines have the opposite gradient (one is the negative of the other).\n
\n
\nIf you sum these two, we get the claimed effect: opinion changes as income varies between $0 to $50k but is unchanged above $50k ... because they cancel out.", "timestamp": 1408716446}, {"author": "Ronald", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673475228122", "anchor": "fb-673475228122", "service": "fb", "text": "Consider the case where 50% of jurors in a particular demographic group would favor the defendant and 50% would favor the plaintiff.", "timestamp": "1408716959"}, {"author": "Ralph", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673475442692", "anchor": "fb-673475442692", "service": "fb", "text": "Have you the link to the calculator?
This may be naive of me but I'd like to think that the question is not so much \"Who will he side with?\" as \"How likely is he to actually understand my case?\". Well educated in a profession that requires rigorous thinking? ... I have a complex but solid case so want her on the jury. Ditzy? ..... I want him on the jury because I have an emotionally strong but intellectually weak case.", "timestamp": "1408717105"}, {"author": "Jeff Kaufman", "source_link": "https://plus.google.com/103013777355236494008", "anchor": "gp-1408718030861", "service": "gp", "text": "@Matt\n\u00a0Given the choice between (A) an unknown person, (B) a person with low income, and (C) a person with high income the defense would (by your model and mine) rank them B>C while the prosecution would rank them C>B. \u00a0They would both put A between B and C because A has even chances of turning out to be either B or C.", "timestamp": 1408718030}, {"author": "Will", "source_link": "https://plus.google.com/115623320734956550423", "anchor": "gp-1408718891667", "service": "gp", "text": "@Jeff Kaufman\n\u00a0By your analogy, you are learning that because the person makes less than 50k you have flipped the \"unfair coin\" and that the odds of favoring the plaintiff are greater than 50:50. By revealing that the person makes more than 50k, you are learning that the coin is fair and that the potential juror has no bias. That is all. There are two separate coins and income greater or less than 50k is what determines which coin you've drawn. The existence of one unfair coin does not suggest that the other coin must be unfair in the other direction.", "timestamp": 1408718891}, {"author": "John", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673478686192", "anchor": "fb-673478686192", "service": "fb", "text": "I was a juror on a complicated homicide case ~22 years ago. There were four defendants for two homicides. Being analytical was very helpful for me and several of the other jurors to understand the case. 6 weeks including about a week of sequestration - 2 1st degrees, 1 accessory, and 1 acquittal. I feel quite confident that we got it right and some remarks by the judge to us after the trial seemed to support it.", "timestamp": "1408719225"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673479474612", "anchor": "fb-673479474612", "service": "fb", "text": "@Tobias: The two possibilities are approximately equally balanced in this case; median income is $50k in the US.", "timestamp": "1408719788"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673479774012", "anchor": "fb-673479774012", "service": "fb", "text": "@Danner: \"Yes, we know that that last spade gave the full house to the opponent, but overall it still looks like the flush in spades is good.\"
But for both of them the last card being different would have decreased their estimate of the quality of their hand. The NYT case is analogous to a person sitting there with four spades treating a fifth spade as good news but any of the three other suits as neutral news. Instead those other suits have to be, together, just as bad as a spade would be good.", "timestamp": "1408719940"}, {"author": "Christopher", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673480821912", "anchor": "fb-673480821912", "service": "fb", "text": "Sure. Except it's likely that the writer meant \"no effect on either lawyer\" to be shorthand for \"if you answer this way it's the same as if the question didn't get asked\". The number of lawyers likely to object to you after this answer (>50k) is the same as the number of lawyers likely to object to you before the question was asked.", "timestamp": "1408720605"}, {"author": "Jeff Kaufman", "source_link": "https://plus.google.com/103013777355236494008", "anchor": "gp-1408720699411", "service": "gp", "text": "@Will\n\u00a0\"By revealing that the person makes more than 50k, you are learning that the coin is fair and that the potential juror has no bias\"\n
\n
\nThe lawyers aren't trying to get someone with no bias either way, they want someone who is as likely to agree with them as they can get. \u00a0So learning \"this person is unbiased (100%)\" is good news compared to \"this person might be unbiased (50%), or they might be biased against me (50%)\".", "timestamp": 1408720699}, {"author": "Zsolt", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673481266022", "anchor": "fb-673481266022", "service": "fb", "text": "Great. \"Do you like crosswords\" -> \"You are intelligent enough to follow the trial\".", "timestamp": "1408720851"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1408720880"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673481430692", "anchor": "fb-673481430692", "service": "fb", "text": "@Elliot: I'm objecting to the NYT's model, which makes claims about probability that don't work out. My model is just \"both sides try to get people who they think will eventually side with them\" and includes \"if X is good news for you then not-X is bad news\".", "timestamp": "1408720999"}, {"author": "Will", "source_link": "https://plus.google.com/115623320734956550423", "anchor": "gp-1408724672646", "service": "gp", "text": "@Jeff Kaufman\n\u00a0We are not given enough information to say whether income under 50k is evidence that the person is unbiased. It could simply be that there is no correlation between the variable \"income less than 50k\" and the outcome of the jurors decision.\n
\n
\nIn the coin analogy, we can rule out which of two coins we've got by observing the other, but here you're arguing that observing a fair coin in one bag should change your expectation for some other coin in a totally different bag.", "timestamp": 1408724672}, {"author": "Ronald", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=673489719082", "anchor": "fb-673489719082", "service": "fb", "text": "You seem to be considering each question in isolation, when what matters is the overall expected bias based on the potential juror's complete demographic profile.", "timestamp": "1408726919"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1408770091"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=675060007212", "anchor": "fb-675060007212", "service": "fb", "text": "@Wang: Your example requires almost all people will answer the question one way. In the NYT example half of people are each above and below $50k because that's the median income.", "timestamp": "1408790630"}, {"author": "Christopher", "source_link": "https://www.facebook.com/jefftk/posts/673465283052?comment_id=675066584032", "anchor": "fb-675066584032", "service": "fb", "text": "I'm wrong in at least two ways. First: some lawyers, even if the income question isn't asked, would look for a proxy for the question. More importantly: given no other information, a lawyer should/would prefer a juror his opponent is neutral to over one they want. I was tempted to argue this effect may be negligible overall, but then the vast majority of lawyers would have to be ignorant of the Conservation of Expected Evidence or use a model very close to the New York Times model.", "timestamp": "1408801142"}]}