Trump and Hindsight Bias
Fraser, Nick and Peter experiment with their own biases, by comparing their pre-US election forecasts with their post-event beliefs.
For more information on Aleph Insights visit our website https://alephinsights.com or to get in touch about our podcast email podcast@alephinsights.com
Transcript
Hello and welcome to the Cognitive Engineering podcast produced by Tell Me Studios for Aleph Insights. In this series of podcasts we take a look at interesting topics and discuss what we think they tell us about analysis and decision making. I'm Fraser McGruer and I'm here with Nick Hare and Peter Coghill of Aleph Insights and this week we're looking at our predictions for the US presidential election and discussing hindsight bias. So Nick, can you lead us in and tell us what are we doing in this podcast?
Speaker B: inspired by an experiment in: Speaker A:afterwards. Yeah so just to sort of make clear so we're now just a few days away from the US election that's when we're recording and we're actually going to record a second half after the election and that is also when we'll release the whole podcast in its entirety. Okay now are we ready to reveal our predictions? Okay so we haven't heard from Peter yet. Do we want to hear what our predictions are based on? Because it's not just predictions it's
Speaker B:probability of outcome. That's exactly what we're talking about. What probability are we putting on the outcomes and why? Right exactly so yeah go for it Peter. So I'm relatively
Speaker C:I'm reasonably confident that Clinton will take it so I'm going to assign a 75% probability of a Clinton win which means that three out of four if we repeated this election over and over and over again in different parallel universes then she'd win three out of four times. Not the same parallel universes okay. Yeah and my general reasoning is that Trump has been having a bit of a slide in the polls recently, lots of bad press. He's weathered the bad press that he's been getting worse so than Clinton even though Clinton, the FBI reopening the investigation into her email server etc she's weathered that slightly better I think than than Trump has and so this I think that he'll take a bigger hit in the in the swing states
Speaker A:than she will. Okay I should explain also that we've all written down our prediction as well.
Speaker B:That's right good practice so that we don't get anchored around what other people have said.
Speaker A:Yeah anything else to add Peter other than that is your coffee nice?
Speaker C:Yeah I think there might be a few surprises in the some of the swing states so the extrapolating from the polls today I think Trump might take Georgia and Ohio but I think Clinton's going to win out in Florida and Colorado. Well hold on then so look let's let's sort of really go to
Speaker A:town on this you've just said I think with some of the states there can you assign a probability
Speaker C:to those states? I've not really thought about that but okay so I think Florida a very narrow win so maybe 60% that Clinton will get it and for Trump in Colorado slightly more I think maybe 70 75% that he'll he'll he'll take that. Okay sorry no not Colorado I mean Georgia 70 75% he'll take that.
Speaker A:Okay Osage, Nick tell us your predictions. I'm considerably less certain than Peter I
Speaker B: utely bang on as it was uh in: Speaker A:measures out there are there indicators that are more reliable like the price of gasoline for
Speaker B:example yeah people never yeah they those those things have been found and they always stop working as soon as they've been found so people do find good predictions so i need to find it there's one that wants to do the super bowl which was absolutely astonishing like for 50 years the super bowl outcome you know predicted whether or not there was going to be a change in the president um and uh and then that stopped working right uh peter you look like you had something to
Speaker C:say yeah well part of my reasoning is that uh trump still got a whole week to stick his foot
Speaker A:in it again right okay um do you have a prediction for what my prediction is i know that you are
Speaker B:more uh i think you're bolder in your in your formation of belief so i think you're going to be higher than peter in favor of clinton winning what do you think peter i think you might be
Speaker C:feeling particularly pessimistic at this time so i think you're going to be bold about a trump win
Speaker A:i predict 100 percent a trump win no no that's just thinking yeah um no okay well look well said nick um so i have a 90 certainty for clinton is what i have and my reasoning is actually very different from yours okay which is in classic fraser mcgruer non-aleph thinking is that i actually i can't be bothered to think about this very much and i just someone said choose a number what do you think and i said i don't know it could be this it could be that there's all these numbers there's all these polls and i just cannot be i mean if i'm just being honest you know i'm very interested in the u.s election and i'm very interested in the outcome but i can't be bothered to sort of to have a proper think about what i think is going to happen and so when we went through this exercise and you know and i didn't even realize that i was going to have to put a number on it and then you revealed to me that i would i just put well i think it's really probably going to happen um so i predict 90 turned it down from 100 just to make it well i nearly went 95 actually um but no that i mean i mean yes if i'm honest that's what i think so lead perhaps just briefly give me we've not gone long left but i mean um as we've discussed many times on on these podcasts my approach is wrong right well we let's i think we should hold that
Speaker B:discussion until after the election okay and then we'll talk about you know forecasting approaches a bit and hindsight bias you know how we so so i think we'll probably have very different views after the election something whatever happens we will if trump wins we'll go oh it's kind of obvious why didn't we see that if clinton wins we'll go yeah that was obviously it was obviously going to win she was ahead all the way we'll we'll find a way to rationalize what happens
Speaker A:and that's what that's what we'll be discussing so in our published podcast what we've just listened to um what our predictions were before the election here we are now um after the election but importantly none of us have listened to um to the predictions that we made so nick can you tell us what you think you said at the time about your predictions yeah well i think
Speaker B:i mean i know i was the least confident uh of all of you um of a hillary victory and i think i said uh i might have said 55 to 60 percent or i might just have said 60 percent um hillary so i was i was of the view that we couldn't draw a great deal from uh opinion polls that they they were you know subject to quite significant hidden swings um and uh so yeah so i i think i i wasn't
Speaker C:hugely surprised by the trump victory peter i think i was more confident i think i was 70 maybe 75 percent hillary victory um and i think i was probably allowing my my sort of uh optimism about a democratic win get it get the better of my judgment um and i i i was also not listening to polls much but i was listening to the the sentiment around the incumbent democratic government and um maybe reading all the liberal press rather than anything else uh but it seemed it seemed like the democrats were were liked more than the more than the republican offering
Speaker A:well i remember with great clarity what i said which was that this was too close to call
Speaker B:and we had to talk i i do remember having to talk you down from 100 percent no no no and very reluctantly i think you said you went down as far as 95 percent
Speaker A:no no i remember saying this is too close to call we shouldn't be surprised if we wake up on on on the morning and there's a donald trump presidency coming now i remember i said yeah i think i said 90 percent and i don't think i was talked down but i think i said you know i
Speaker B:i thought about 95 percent um i i think you started out by saying well obviously hillary's going to win the the use of percentage percentages here is is is absurd because she's obviously going to win no i didn't say that okay no i think if we find out when we listen to it that you did say that what what do you what will the consequences be for me well i mean just no but i think it's this is really interesting because i i actually have a distinct memory of that but you don't
Speaker A:well no one of us is going to be okay so okay good point so what i think i said was that um i was fairly certain that uh clinton was going to win and i put 90 on it 90 percent certainty um i had been thinking about 95 or 100 but you know i'm i'm i'm i was i'm more cautious than than that and so i said no 90 but i also remember why i said it um and i said look to be frank i didn't realize we were going to give a percentage before we started broadcasting so i just thought one up and i can't be bothered to think i've got better things to be doing you only know two percentages 100 or zero thank you and um i um i just couldn't be bothered it's just too much thinking that's what i said um i think it's interesting the my the extent to which i was certain and the and the my rationale behind that is was very wrong um i don't know i think i'm the point i'm trying to make is that even though we were all wrong is that i was the most wrong and according to aleph methods the the the method that i was using to get my decision was completely wrong and inadvisable and sure enough
Speaker C:it was um i don't think it's just aleph methods i think it's like the scientific method yeah yeah
Speaker B:it's sort of doing doing proper forecasting you know i think we touched on it at the time but we certainly touched on it at other times which is trying to trying to put the situation you're in into a larger reference class of similar situations and we really didn't have to go back very far in the year to find a similar situation um where you know that it looked like a fairly close call but that polls were putting um remain uh ahead of the brexit um and of course it was the other way around and and here we were now i know people have drawn um analogies between brexit and trump and some of them are valid but just even forget about that just think of it as an opinion poll where uh you know the where the the polls are showing a slight edge for one side um you know forget about whether there's similarities between the kinds of underlying motivations for trump voters or brexiters you know polls are often wrong and and you know for reasons that are hidden to the pollsters um and this was another one of those situations it was very close the opinion polls were fairly close and um you know we we shouldn't i mean and this is the thing that i think we can look back now and most people probably are telling themselves that they weren't really as surprised as they were and that's really why we did this experiment was uh to look at this phenomenon of hindsight bias which is phenomenally important um for all sorts of reasons and it's this it's this phenomenon of um constructing a narrative explanation for things that happen to make it seem like they were sort of more inevitable than they were it's sometimes called the i knew it all along effect and uh it's been heavily researched and i think one of the one of the one of the things that hindsight bias does is it makes us it makes us forget that we were surprised and so it means that it means that we um actually go through life thinking we were more right about stuff than we were and when you actually go back and investigate your your your past beliefs you can find that they're actually quite different you can be quite surprised if you've ever read an old diary that you've written you could be quite surprised by the things you thought um and there's an interesting question as to why that happens so why does it happen well that's no no one's quite sure the the question is really whether it's whether it's adaptive so do is is the reason that we uh revise our beliefs about what we used to think is that is that some sort of engineering trade-off do we do it because it's just you know our brains find it easier to operate that way or is there some storage for yeah we used to think or is it actually that there's some evolutionary benefit to forgetting about what you used to think well are we not talking about cognitive dissonance here go on are you familiar with this well yeah i know i know the term okay so yeah i mean so the idea is that
Speaker A:we can it allows us to hold two contradictory views but sort of settle on one of them and go that's that's what i was thinking all along or i yeah i feel okay about that looking psychologically that's useful for us otherwise we'd it would just drive us mad as the most wrong of all of us
Speaker B:looking back uh do you feel i mean what do you think about old past fraser there from a month ago uh what did what do you think i mean do you think what nincompoop why was he saying no do
Speaker A:you know what i think um in all seriousness it makes me think that i have a greater uh craving for certainty than either of you two do um in in in anything really i i don't really have patience or time for things that are more ambiguous um and i just recognize that as a trait in myself just because i recognize it doesn't necessarily mean i'm happy about it or i couldn't try to
Speaker B:change it well you can take comfort in the fact that it's not just you it's also a lot of governments um and you know thinking about the british government there was certainly no plan b for for what would happen if the brexit vote voted to leave um and and you know why why not i mean obviously everyone was just absolutely making it up as they went along on the next morning you think what that's extraordinary but really deep down people didn't think it was and a similar thing with the trump vote you know actually you can see that a lot of international governments were wrong-footed by it and don't really know you know they don't have a trump strategy that they don't know who the key people are you know why why are non-entities like nigel farage suddenly being propelled into the limelight because they're the ones who've been keeping up contact with trump and his team in a way that um you know we actually our officials
Speaker A:i think sorry i know you want to come in peter but i just also just want to say i think the other thing is as well as this craving for certainty i think i'm a more emotional person than either of you two and and so the thought of a trump presidency was such anathema that i just couldn't contemplate and i didn't want to think that that could happen and seriously
Speaker B:how could anyone yeah but why i mean this is the problem with hindsight i mean this is the puzzle with hindsight bias it cannot serve us to believe things because we want them to be true it must be more advantageous to believe or attach the correct probability to things so that you can make the right decision about them of course you know believing things that we want to be true doesn't ultimately stop you know the the crocodile biting i recognize the frailty of this position yeah
Speaker C:peter so aside from the fact that governments don't make enough resource available for planning on contingencies i know for a fact that some government departments weren't allowed by their ministers and civil servants to senior civil servants to plan for the uh scottish the outcome of the scottish referendum should it have should it they they have voted yes to leave the union um that so it's seen as a waste of resource but aside from that a lot of our customers that we've known from uh ministry of defense they they are unable to accommodate a probabilistic judgment on an outcome they they they are all they they're all they're after certainty most of the time they want to know will this country invade this other country or is it yes or no give me the answer is it is and there's a sort of naive perception that there is an answer out there that can be known um and that it's the job the analyst role to find that answer out not to place a probability on that outcome which as nick says is more useful and anyone who does risk analysis will will understand the concept that you know if it could go one way or the other and you can work out what the percentage is and you know what the impact of each of those outcomes is that's a very useful um set of data to work out what the your expected cost or value to that that that that ultimate
Speaker B:outcome is anything else out here just um you know we talked about what why what hindsight bias is and we should probably point out you know obviously this is not in any sense a proper controlled experiment this is really just uh experimenting on ourselves subjectively to see what it was like paying attention to the to the change in uh you know in in um our own beliefs about about our own beliefs and i noticed even just you know as soon as the vote had happened i remember trying to remember what i'd said and convincing myself that i'd said something lower than i had i think what i said was 60 percent but i remember convincing myself that actually what i'd said was 55 percent i could see that revision happening in real time and i think the there isn't a consensus about why hindsight bias happens but um the to my mind the most convincing theories is simply that it's a way of conserving um you know cognitive uh cognitive space um why would you hold on to incorrect old information the idea is that when you learn new information so the new information we might have gained from the american election is that trump was more appealing than we thought once you learn that why why bother retaining a cached version of your past sets of information you know you don't go through life uh collecting um old versions of your own beliefs and putting them somewhere there's just no point we have not evolved the capacity to do that um yet you know hindsight bias actually is a reflection of the fact we're updating our beliefs so that when we look back and say well what what actually how likely was trump to win um we're taking on board the information which actually we didn't have access to which was turns out he's quite inexplicably to me quite appealing well look but here's the thing because
Speaker A:um and i want to wrap up here but actually i'm the victor in this okay and the reason right yeah and the reason why is there there's there's a certain sense of irony here that um my uh decision making um methodology was was poor and it was let's call it shooting from the hip going with my gut going with their my emotions however that sort of methodology is not the sort of methodology that will prevent you from being elected as president of the most powerful country in the world and um people will ignore all you smart analysts and they'll still go with someone like me so um or donald trump although i'm slightly concerned about putting myself in the same stable as trump and um all i could say is all hail president fraser here i come peter that is an interesting
Speaker C:thing because the the famous pundits the successful pundits aren't and as ted lockett has demonstrated are rarely the the most correct the one they're just lucky uh or they're the most vocal about a particular subject that that people like to read about so um yeah there is something in that and the the the analysts who are really bayesian like ourselves um will will not meet fame and fortune because we'll be written out by the pundits who bang on about um water wars or or uh or some
Speaker B:other sort of peculiar topic yeah no no one's gonna no one wants to listen to some guy come on the news and say well you know i feel very strongly that this could go either way you know they want they want to have they want to have someone who's going to give a really convincing story as to why one thing's going to happen and then have another guy who says the exact opposite yeah well that's
Speaker A:why in the us um and i guess elsewhere is the amongst political pundits there's lots of right wingers out there is they get a lot of viewership or listening or listening audience because it's it's more interesting to listen to someone sort of ranting and it's not just right wingers i mean
Speaker B:this is this is one of the problems probably safe for another podcast but you know that people are listening to the news that supports what they think and that's certainly true of liberals as
Speaker A:well absolutely no there's definitely a podcast there okay um look so um let's wrap it up there um as i said you know just to sort of reiterate all hail president fraser here i go um thank you nick hare thank you peter coghill um thank you president fraser uh you've been listening to the cognitive engineering podcast with uh aleph insights thank you for listening and until next time bye
