Peter Donnelly: How juries are fooled by statistics

[rafelingd]

he reminds me of rupert giles from buffy.

[Triipticz]

5:25 failed joke

[ThePlatoon4]

when it comes down it, its really just being persuasive :p

[EmergingCurrent]

99% of all statistics are made up on the spot...

...This is the 1%.

[taantumus]

Quite interesting really, but the title is more or less misleading.

[andreeaweed]

this is very interesting course about statistics ...i never thought about that in this way

[TheNishnish1]

who gave a pediatrician the right to make statistical decisions in the first place. Dont ppl know pediatrician = doctor = knows medical related stuff and not statistical stuff. Giving a pediatrician authority to make statistical judgments on a case just because a child is involved is like allowing randy jackson to judge a dance show just because both fall under the umbrella of "entertainment" . A woman lost her life because of this mistake and its really sad.

[47f0]

It's not just statistics that juries aren't clued in to. My ex, a reporter who had the court beat, watched an innocent man go to prison for robbing a bank - the jury simply couldn't understand the simple pythagorean geometry, that, knowing where the bank robber was standing, and the height of the surveillance camera, that the defendant was a full six inches taller than the actual bank robber. Worse, her editor nixed any explanation of what went wrong - he said "nobody would get it".

[penneyduk]

Hey, this looks like it's from the same session as Steven Levitt's talk on children's car seat safety. He's an economist but I'm sure guys will find it interesting as well, it's based on statistics.

/watch?v=um5gMZcZWm0

[ByteSmasher]

Best line in this: "you keep tossing to wait for the next head to get excited" >.>

[DontTouchMyVicodin]

He models jeans. xD

[pakjenal]

i got the disease and the jury examples..but im still confused about the coin toss example. anyone care to explain? thanks :D

[OneSummerSky]

82.7% of all statistics are made up on the spot.

[obliviousaa]

gets good at 11:00

[guyboy625]

@obliviousaa That would be missing an important part of the point.

[eldadevata]

I find the system based on a jury mediaeval to say the least. A strange by-product of a time when the “community” was responsible for judging its members, without any expertise required. It might be that during a limited era, this system helped some people at getting a relatively fair trial under an oppressive government, but today I believe its obsolete. I would never trust my life in the hands of a number of people who are not trained to think rationally and weigh the evidence.

[Grawp55]

@eldadevata ...but if you know you are guilty, a jury gives you a great chance to FOOL them! Much better chance of getting out of there than with a judge!

[eldadevata]

@Grawp55 Yep!

[onehairybuddha]

@eldadevata

Unfortunately judges and magistrates aren't any better, at least with a jury you could get one clear thinker who might have an effect.

[libertarianjury]

Excellent. Of course, the perverse incentives purposefully caused by all other "systems" are worse than random juries. And, juries themselves are no longer proper juries. "Voir dire" is a new arrival (1850 in the USA), that lets the prosecutor pick juries that are biased in favor of the law, for instance. The same with the licensing of lawyers under the "BAR". Add false judicial instruction (1895) to that, and 'contempt of court', and juries are anything but "randomly inserted judgment".

[eldadevata]

@libertarianjury I cannot imagine living in a culture where people are picked randomly or based on their bias, to pronounce a verdict, without having any expertise in judging the evidence presented to them in a logical and rational way. It might be that the continental law has its flaws but I do prefer it highly above this circus. Of course one judge can be limited, so we designed our higher appeal courts to include several judges.

[CalicoVall]

My fucking hero.

[BenThomasFoster]

The statistic videos are always the best once!

[girasol1111]

I am writing a paper

this video was quite helpful.

Thank you, Peter Donnelly you are the

James Bond of STATS.

[ninjatoothpaste]

I've been thinking about head and tail all day and I've tossed so much my hand aches...

[simonwedege]

When I saw the word "statistics", I thought it was wrong to be dull. But it certainly one of the best TEDtalk I have seen. Really important topic as well.

[cryora]

So basically the pediatrician made the mistake of assuming that the chances of a second child dying from this disease is unaffected by the chances of the first child dying from the disease. Yes that is a horrible mistake. Assuming that the first child has the disease, the chances of the second child of having the disease is much higher, because they are children of the same household and parent.

[statwiki]

nice points

[unsatura]

wow, a bang on the head!!

[LottsaLasagna]

Just tell them you're a Liar..Problem solved.

[bananian]

just have several statisticians do an analysis independently and see if they match up and have them explain the analysis as well.

[hotkonto]

In my opinion, one of the best TED talks. Wish they had more talks about statistics and probability.

[Vegie007]

Statistics are to a lawyer what street lamps are to a drunk. They they lean on them without getting much illumination.

I'd rather have this guy on my side at the trial.

[TejasM14]

What an insightful analogy!

[arash3589]

beautiful!

[MrJakedonnelly]

My name is Zach Donnelly, we have the same last name. Thats pretty cool!!!!!!!

[astat1]

Yeah, that's TERRIFIC!!!

[Kotesu]

Daniel Jackson!

[richardcadbury]

Guess what happened to the expert.

He had his medical licence revoked.

Then he appealed and a court gave him back his medical licence.

It's a scandal, really.

[sparkloweb]

The pediatrician's lawyer hired an expert dendroclimatologist to argue that there was only a 1 in 73,000,000 chance that someone could be that stupid, so they had to give his license back.

[cosmeticnerd]

i love statistics, too bad i am horrible at math. after i took my stats clas, it changed my entire outlook & realized how ignorant people when they analyze the data.

[Dayvit78]

he took way to long to get to the subject of the talk, and also his jokes are not that funny. they were ok at the start, but at some point, just stop, and try to make what you're talking about interesting instead.

[Xiph1980]

About the funny-ness of the jokes, well, that's a personal opinion, but I do agree that the intro should've been a bit shorter. Remove out the "other person's shoe" etc. But.... The presentation is extremely interesting in my opinion :)

[franciscosfcabral]

statistics are precise unlike what vjpillay says. What is not precise is our interpretation and the way we deal with it most of the times

[Chasebald]

Uber-nerd.....!

[sidexwalk]

He's an oxford mathematician. Did you expect him to be down? Hip? Cool flow with a funky jive? No dude, he's a fucking Oxford professor. This guy deserves so much respect it's hardly funny.

[lasttarget]

@sidexwalk very true; thank you. for your comment.

[BenThomasFoster]

@sidexwalk why does being a professor at a uni command respect? genuine question. sure if he solved blah blah blah thats not theoretical but real then yes maybe but as far as i'm concerned he's just clever

[sidexwalk]

@BenThomasFoster at one point, Peter Donnelly was the youngest uni professor in great britain, his work is the backbone of the theory of coellescence, and he heads the statistics department at Oxford. need i really say more??

[BenThomasFoster]

@sidexwalk yea say that he's save some ones life. say he volunteers to do care work in an orphanage... I could care less if he's made a rich person richer because his ric family could give him a private school up bringing

[sidexwalk]

@BenThomasFoster it takes more than a "private school upbringing" to head an acedemic field at a prestigious university. you do realize that requires a devotion that ranges beyond whether he was publicly educated or not. if you lack respect for academics then go learn how to spot-weld and stop engaging in intellectual discourse.

[fuckooo]

@BenThomasFoster You're an idiot.

[muzvit]

Uber cool...!

[tausendstein]

Example:

Beck-Bornholdt, Hans-Peter and Hans-Hermann Dubben (2001): "Der Schein der Weisen - Irrtümer und Fehlurteile im täglichen Denken. [The illusiveness of the wise men - Falsity and misjudgement in daily life].

[tausendstein]

its not a bad speech, but it is a somehow 'old hat' for statistician what he is talking about. And there are also many 'for public use' written books about such effects of statistics and errors by interpreting statistical results.

[mlawren7]

If you were a defence lawyer and a expert witness said the chance that your client is innocent is 1 in 73 million, wouldn't you investigate this claim a little further?

[Xiph1980]

Most wouldn't. Infact, most wouldn't even know where to begin to investigate that claim.

[vjpillay]

If a doctor says 60 % cure rates means on average 60 percent of people with the particular cancer has been cured in the past and so be positive. If one is positive, 60 can become 70 % and on he way to cure . If negative, 60 can become 50 %. It all depends on how the patient deal with hope for better. So be positive..Doctor is not god but help one to get better but all down to patient to remain optimistic. I hope i have enlightend since life is precious.Statistics is not precise mathematics

[benwaddelling]

Oooo. Big wordz. Me not no what think.

[richardscientia]

In the case of Sally Clark, statistics can not tell whether she is or not innocent, but the chance is statistical like any other type of Sally Clark case. This is where statistics goes over its limitations. You can not generalize a special case, because it is special in every way. Special cases are not in any way general, but take some special cases, you can make a general understanding by those with the tool called statistics.

[Erudecorp]

No, that is further ignorance, since that is what most of Statistics is about. Professor Donnelly, the speaker, a Statistician, explained that the Genome project studied what we have in common, and that the new study is about what is different. Description tells you what's the same about a group of data. Inference tells you what is different or special. There even does exist a specific definition for outliers. Sadly, few people, for instance you, actually bother to study Statistics.

[richardscientia]

Two Sides. Statistics will always demand a group of data to prove something. On the other side, Sally Clark case is juridical a special case and doesn't belong to the statical view. If hard evidence prove Sallys case, it will be regarded as special evidence, not general evidence. I believe it is ignorant to judge anyone in particular for something that comes from generalities. One day U may stand there marrying your partner, and u say -I only take u because u r not special, but very general :)

[richardscientia]

Still, there can not be any statistical evidence against Sally Clark. You need hard evidence to prove a case of singular action contained. Any definition regarding outliers does have its reference to the general data. It contributes to understanding, but not a conviction.

[CablloBlanco]

Outstanding presentation

[JocularINC]

This seems like rather simple logic.

hthththth = 9 letters and 4 sets

htthtthtthtt = 12 letters and 4 sets

The likely hood of hth apearing in a random string more often then htt is statistically more probable simply because, like he said, hth overlaps like in the example above.

I think he does not explain himself so well which is what may be causing confusion here.

[soccom8341576]

Interesting talk. I happen to be doing a statistical project right now.

The jury system would not be as bad as to get everything upside down wrong, if some statistician was also used as an expert.

Everyone makes mistakes, and these mistakes are encouraged because we have trust in authority, so that when authority gets it wrong, we get it wrong too.

I think statistics is beautiful in giving a general impression of fact, but whenever possible, not to rely on it.

[L02T]

What else is there to rely on if not statistics? Arguments?? Those are subjected to even more bias and errors compared to statistics.

I would say, when it comes to analyzing a certain data, do not rely on only 1 person. Rely on a group of people. A team. Any errors made can be overcome with some proofreading. Not fool proof, but still more trusworthy than anything non-statistical.

[soccom8341576]

well, more logic could be helpful. possibly analysis of psychology by a team of professionals, with a wide range of knowledge could help the jurers.

It may help reduce the potential for error, as it is obvious that statistics are quite good at triggering emotion.

Some logical evidence to back up and explain the conclusions from stats could be used.

I guess I am a little weary of statistics.

[dpsadin]

he has such a big head and he's so full of himself.

[ferrisbueller9000]

Yeah,,,you're an expert

[jabmeister]

Superb video.

[George4943]

In a single flip-stream race, it is 50:50 as to whether the HTT-seeker or HTH-seeker will "win."

However if the HTT-seeker and HTH-seeker uses a different flip-stream, the HTT seeker is likely to find his first. HTT first-occurrence-average = positions 6-7-8. HTH first occurrence-average = positions 8-9-10.

[molewizard]

My statistics teacher did basically this lecture for a whole lesson.

He was so mind-numbingly dull. This guy isn't.

I hate my statistics teacher.

[George4943]

I ran the simulation.

The experiment is to flip until either HTT or HTH appeared. When that experiment was simulated 1,000,000 times the average number needed to find a "winner" was 4.99. And it was 50:50 (NOT 6:8 as claimed) as to which "won" the race.

I will be glad to provide the visual basic code.

[George4943]

Replying to self can be fun.

His statistic is accurate. I ran a different simulation and indeed the average n at which HTT appears is 8 and HTH at 10.

[molewizard]

Did you inlude this:

HTHTH

As two HTHs?

[melly021]

quite possibly the worst clip on youtube!!!

[forestskog]

Excellent. Thanks for posting

[dorotwhy]

So... what does it mean when a doctor says, "Your cancer has a 60% cure rate."? Never mind, I think I might not like the answer.

[dorotwhy]

Watching this video reminded me of my inadequacies in mathematics.

[patnoon314]

hes so right, human mind is really good at some thinigs but often dead wrong on others

[vjpillay]

I am surprised the statistics expert was not called instead of relying on paediatrician,who would have got his statistical interpretation wrong? nest time,if a doctor says statistically,one has to ask if ti isa backed by a statisitican?

[Erudecorp]

Because those judges and lawyers have the educational level of a social worker.

[TheMathGuy]

That's rather scary to think about! I wonder how many innocent people have been convicted because of errors in statistical reasoning.

[Truthiness231]

Yeah no joke. I was "100% certain" in my mind that B was the right choice... just seemed to make since that once you have "HT" that there is a 50% chance for a "H" and 50% chance for a "T" (to complete "HTH" and "HTT" respectively), but yeah I didn't take into account the restart of the pattern if you don't get the pattern you're looking for... spooky how our minds don't work when in comes to statistics. And that we fall for the appeal to authority fallacy when it comes to statistics too...

[nerdyharry]

its funny, drummers would know intuitively from doing paradiddle exercises and rolls that it takes longer to establish a triplet pattern with one consecutive repeat than one that alternates.

[waylay00]

Interesting video.

Reppin' MUS Period 3 ACC Precal

[marcarmstrong88]

i completly get wot he's saying but i understand nothin about it.. as in i can see he's completely right but its hard to get my head over it!

[popeyroach]

awesome talk

[CrispyClaire]

Wow, this was a great talk! I'm not into numbers much, but he put it so it was interesting and easy to follow. Very, very good lecture.

[caughtbytheflash]

omg! thats my friend's Dad, no jks!!!

[papasitoman]

60% of the time it works everytime.

[Jakemmx]

I think the whole Jury system is quite flawed.... It should depend on critical thinking and evidence, but it seems to depend more on attorney talent and the emotions of the jury.

[BoozyBeggar]

Plus each individual juror can affect the opinion of other jurors, or one juror could rise to 'alpha juror' status and be backed by 'beta jurors' and make an honest dissenting opinion of another juror more and more likely to vanish.

My choice of language and metaphors were a bit fuzzy, but hopefully you get the message.

[tehmushroom]

one word..the runaway jury

okay, a phrase.

[xXxnerdlovexXx]

Not to mention how much money each client has.

[jazztenor71]

Aw, people can come up with statistics to prove anything. 14 percent of all people know that.

[DrewJ16]

to prove and/or disprove.

[memblor]

it's 'forfty' percent of all people know that.

[cryora]

lol at the end there was car commercial

[WolfDarrigan]

because they sponsor TED.... its on the end of all TEDtalks

[cryora]

"We are not good at reasoning with uncertainties"

[JimGrimReaper]

Are you sure ?

[JimGrimReaper]

That we aren't good at dealing with uncertainties, that is.

[Efferts]

You say that as if we shouldn't be punished for it.

[JimGrimReaper]

I'm not sure that we should, but then again...

[intrepgun]

the universe is full of chaos. and chaos is basically good. the human's inability- and i believe it is an artificial inability, grown out of some of the sillier parts of society- to deal with chaos and change will only breed problems. people need to relearn how to cope with life and float on the chaos.

[JimGrimReaper]

They say that chaos is just order that we don't yet understand....

One day all of this will make sense, but that day will be long after we here are all gone.

I'm sur of that.

[intrepgun]

i think that's a real possibility, but as we don't yet understand a greater order [outside of the feeling of being part of the 'great work' and a step in the evolution of everything] i think we're more animal than we understand, and should try harder to live in balance with our environment while taking steps to grow beyond the planet and set a goal that humanity can get behind.

[shsnawada]

Wasnt that bayes theorem that he was talking about when saying 0.00098 and all that?

A very good talk, btw.

[emperoratuni]

yes that is bayes theorem exactly

[Flyrev]

Sorry, limitations of space here: [...] is higher in the HTT example. Because when you get HTH, you don't have the HTT pattern, but you only need two more to get it. Understand?

[Flyrev]

You're wrong. A is correct. Just because the HTH can appear in clumps, doesn't mean that the average number of tosses before the pattern appears is higher.

[JoseEmidio]

Yes I think you are right.

[rksinc]

In his early example tossing coins, he says A) is the right answer, but then proceeds to prove that C) is the right answer. His explanation is clear: throwing HTHTH gives 2 chances to get HTH, while there is no way to throw 5 coins and get HTT twice. Therefore, C is true, not A.

[Doc1806]

No, i believe you are wrong.

[lefthandrighthand]

Nopes, listen again starting at 6:50.

[hoschiadedodi]

Well I think he was explaining that when you throw HT and you want HTT but you get HTH you already have the H for your next Trial (on the way to HTHTT). On the other hand, when you throw HT, looking for HTH, but getting HTT you have to start with an H again (on the way to HTTHTH, which is 6 cyphers)

[throwbakman]

if it hapensmore times in 5 coins, then it will be larger, not smaller then HTT .. not gunna explaine it so just look at it and think till you understand that it is A

[happypeacock]

fascinating. I learnt a lot!!

[rafaelvongehlen]

Wonderful video. Thanks for posting!