Unraveling the Mystery of the "Curse of Knowledge" with Steven Pinker!

well hello and welcome I'm your host Sam Knowles and joining us today is Stephen Pinker the Johnson family professor of psychology at Harvard University an experimental psychologist interested in all aspects of language the mind and human nature Steve is one of the most important public intellectuals and best-selling authors of the past 30 years welcome to Global attention with this 1994 book The Language Instinct and followed that three years later with how the mind works I first saw him speak indeed just after that book came out in a debate with Richard Dawkins titled is science killing the soul appropriately enough held at Methodist Central Hall in London in the 2000s Steve's interests and Popular Science bestsellers have flexed and grown to cover human progress the prevalence of violence or otherwise in society and rationality many listeners will be familiar with the blank slate the better angels of our nature Enlightenment now and most recently rationality one of my personal favorites is Steve's 2014 book the sense of style the thinking Person's Guide to writing in the 21st century and what it has to say about the curse of knowledge we may come back to that a little bit later garlanded by media national International organizations and academic institutions around the world CBS generally agreed to be one of the world's leading thinkers and most influential writers he is that rarest of creatures a serious practicing academic who writes with great clarity for both his peers and an intelligent layered audience and we are honored to have him join us here today Steve first of all welcome to the data Malarkey podcast thanks for having me Sam now I trust that's a decentish summary of who you are and what you do is there anything you'd like to add anything I've got wrong it was so complimentary and flattering that I I feel uncomfortable either accepting it or rejecting it but thank you for the kind introduction and uh I'm a I guess you could add in the context of this particular uh show that I'm a big fan of data very good very good now listen before we get down and dirty and focus on data I wanted to go a little bit beyond the CV and start with the question designed to get to the heart of you it doesn't necessarily ask us to tell you about your job so could you tell me Steve how do you spend your time okay I I yeah I do a lot of reading I I teach I write I have a day job at Harvard University when I'm working on a book I tend to immerse myself in it and I once I start something I have to complete it and so I tend to work seven days a week but otherwise I try to experience the rest of Life by with photography with bicycling kayaking uh hiking um my other half Rebecca Goldstein is a novelist and philosopher and we have many interests in common including walking in cities and discussing ideas as well as far more mundane mundane matters to concern any couple in a marriage now you mentioned photography um I know that that picture behind you of the of the Boston and Cambridge waterways is one of yours could you tell us a little bit about your passion for photography you have an extraordinary array of pictures beautiful pictures on your personal website yes I've uh partly it's like like a lot of guys I like equipment and like I think lenses and photograph equivalent are beautiful and fascinating it but it also comes from my interest in visual cognition that's how I began in in my graduate career studying imagery and attention and recognition of shapes and themes I'm interested in the Aesthetics the visual Aesthetics what makes a landscape beautiful what makes a face beautiful what makes a a shape beautiful and so it's a nice way of combining some of my intellectual scientific interests with a a personal uh a hobby and an opportunity to acquire some cool gear excellent well I was I was flicking through before a couple of nights ago and as a having family Connections in New England I was pleased to see a good number of excellent Red Sox pitchers in there as well Fenway features large it does indeed now can you tell me um can you tell me that you talking to me about yourself as being a fan of data can you tell me the the role that data has played in your career and how this has maybe changed over the course of that career yeah certainly the one of the uh appeals of cognitive psychology When I Was An undergraduate and floundering for a major was that it it promised to take on some of the Big Ideas in in human history human human nature how do we know the world how do we what's the nature of language uh what kind of creatures are we but take them out of the the seminar room the the armchair and at least hold out the promise of making progress and understanding through empirical testing through data uh and early in my undergraduate career I remember the first time that I uh did a really serious experiment I was working in a lab on auditory perception um I in those days using I was ahead of my time even in using computers because many people just use calculators and you get a wide tractor feed uh print out you know in in all caps of your data and the computers weren't even smart enough to convert a matrix of numbers into a graph and so I had graph paper and new colored pencils and I remember just plotting the points reading them out from the computer printout and they kind of fell into this beautiful pattern exactly as predicted by the theory and that that was a kind of moment of ecstasy that just cemented in me the desire to be someone who uses data the idea that that the prospect that an abstract idea whether you were right or wrong could actually be seen in shapes and colors uh was it was was truly exhilarating and and ever since then I've taken great pleasure in especially when there's a difficult um conceptualization problem you're not sure what your data are telling you you've got Matrix and tables of numbers and then you figure out how to plot it in a way that shows what's going on that kind of epiphany is a moment of of uh of deep pleasure but on top of that so that's just my background that's just why you know why I'm not a philosopher why I'm not an artificial intelligence researcher data is part of the pleasure of of what I do but also and this connects to my two of my more recent books Enlightenment now and the better angels of our nature it can change your view of the world or your view of human human Destiny and in particular in my case uh seeing um graphs of measures of human well-being plotted out over time and seeing a number of bad things like violence going down now you and I had a discussion as to what was the orientation of the my image on the the uh the screen because it's gone through several layers of processing so I don't know which is left and which is right yes am I is this going uh the right direction from a graph from left to right uh it is going having time having time is going that way that is definitely going there I'm seeing it mirror reverse and that's that's what I ask anyway there are lots of graphs that look like that bad things going down good things going up uh violence uh child mortality maternal mortality number of countries with slavery racism uh child abuse domestic violence graph after graph when you plot it it's going in in a good direction uh this was for me mind-blowing because um for one thing it overturns the kind of impression of the world that you get through uh journalism and through history uh which is since it's those are a non-random sample of the worst things that are happening now or the worst things that happened in the past the wars the revolutions the terrorist attacks the uh the epidemics uh as long as those haven't gone to zero there are plenty of them to fill your news feed or your history book whereas all the things that don't happen like countries that aren't fighting Wars or don't get attacked by terrorists or the things that build up incrementally a few percentage points a year which then compound are just invisible unless you're looking at data and that matters in terms of you know what our species done have we just got made it bigger and bigger and bigger mess or if we actually pushed back against the forces of the universe to grind us down so that's a big Cosmic philosophical uh understanding of you know the cosmos in our place in it and for me it it came from data it couldn't have come from any other place yes man doesn't buy dog rarely sells a a copy of the Boston Globe doesn't it that's indeed in in terms of your your professional work I I mean I I I I I I I've read um almost all of your books including um uh words and rules and uh um and I'm very interested um in the types of data you talk about you talk about those particularly the types of data in terms of violence um uh and and progress and health um I'm very taken by the by the work of the late the Hans rosling and the Gap minder Institute um and the way that they too bring sort of a great sense of optimism to the world when when people working in public health and epidemiology may be saying we have you know all of these challenges so I'm taken by that could you talk to us a little bit about about the types of data um in the research that you've done over your career that have meant most to you be beyond the ones from those two books um well I've um I you know I I'm trained as an experimentalist but it is frustrating to do your own experiments because you can only kind of dribble in a few uh observations at a time whereas the bigger the data the more solid your conclusions so I've been you know I started out I I would bring college students into the lab and offer them some you know beer money to sit through an experiment and they would see shapes and press buttons or see words or sentences I'd bring kids in uh or go to daycare centers a little duck and Pony puppets and you know and that was good but um we are coming to appreciate more and more that our intuitions about how many uh subjects you need for a sound conclusion are likely to be faulty we should have known this a long time ago because a mistersky published a brilliant paper in 1970 called The Law of small number belief in the law of small numbers that was a kind of nerdy joke at play on the law of large numbers the law of small numbers being not a lot of statistics of course but a law of psychology which is that we humans tend to believe their sample of any size is going to be representative of the population which is drawn so we tend to underpower our studies we tend to uh one of the reasons behind the notorious replicability failures um the uh and so I've been heartened you know growing up in the in the tradition of running running you know 10 15 20 you know maybe 30 subjects uh the revolution that allows you to sample data sets that uh are the fruits of many other people's research in the case of child language acquisition which one was the other of my early interests in Psychology in the mid 80s in a kind of precocious um Big Data moment uh a number of um my fellow psycholinguists digitized transcripts of children in conversation with their parents sample say an hour a week over several years and uh and and um amassed one of my graduate advisors LED one of those studies in the 60s by Poe he studied three kids uh he was ahead of his time but his data together with people from all who will got Amalgamated into something called the child language data exchange system and that that revolutionized my life in the 19 back in the 1980s when instead of 25 kids playing with ducks and bunnies I could have you know tens of thousands of sentences and I published a number of papers of uh analyzing that the the trends in uh children's speech even then um it was I was hungry for more and I actually wrote a grant proposal or something that we whimsically called the human speech on Project which is to try to get instead of an hour a week to try to get like everything that a small number of children said and that their parents said to them that is at least an order of magnitude more data the it was turned down I didn't get the money for it and I I did other things but that that was my my dream at the time that was one area in which data uh just kind of changed my life although more would have been better still uh also analyzes and this again was kind of precocious before the Big Data revolution of the 21st century one of the first thing sources of Big Data were a corpora of a written language uh there was something called the uh Brown Corpus led by cutra Francis a million words of English text and I went back to that many times in looking at just the structure of the English language the statistical structure just to give you an example just so we're not being totally abstract I was interested in regularity and irregularity in linguistic phenomena so we've got you know from the past tense in English you've got walk walked and explained explained and uh talk talked and uh but we also have Irregulars break broke run Ren come Kane uh which are have to be memorized because they're idiosyncratic well here's an interesting fact if you look at a data set of the English language in order a frequency of words how many times does a word occur uh per million words of text you find that the most frequent verbs in English are irregular uh they're a minority of word types they're only about 165 irregular verbs compared to about four thousand verbs in all but they are the top 10 in English and they probably make up a majority of the actual usages that's an interesting fact and I think the explanation is that irregular forms like concave or go went sing-sang they got to be memorized because there's no rule behind them they're exceptions memory can only cope with things that are repeated often enough to uh actually be Consolidated in memory and so there's a constantly a darwinian process in the language itself where words that are frequent enough can tolerate irregularity because people hear them often enough that they can memorize them when a word declines in in in frequency uh like child whose past tense used to be chit then you got a generation of children who won't be exposed to he cheat me often enough for it to stick in memory it'll then default to the regular side and you can try it so that's another example of a kind of epiphany that could come from looking at data on language but that's in my the phase of my life where I concentrated on language and visual cognition and then as I came across data sets on war and and democracy and hunger there were a whole other set of epiphanies I'm really interested are you mentioning the replicability crisis would you say that um the uh you know one of the problems being the use of weird subjects the white educated industrialized Rich Democratic subjects um that so many psychological experiments have been done uh using would you say that the the Big Data Revolution is starting to to eat into the anxieties about the replicability crisis is that helping to mitigate some of that well I think the the Restriction of of uh research to uh modern industrialized Democratic subjects is itself a problem I don't think it's the cause of the replicability crisis because the replicability crisis was that you do the same experiment with more weird subjects and you know replicate uh if you were beginning to try it on um you know uh Indian peasants or or residents of the Bolivian Andes that's a whole other potential source of replicability failures but the crisis itself was you just get more college students and it doesn't replicate they're they're I think there are a number of causes and they talk about them in rationality one of them being the underpowering of research that Amos trusky to no avail warned us about in 1970 that is the cognitive Quirk of thinking that a sample is going to be representative uh but there are others um post-hoc hanky-panky and data analysis uh again we really should have known better because I remember as an undergraduate in the 70s my stats Professor talking about what do you call he called it at the time data snooping uh that is you have your data and then you choose which statistic statistical test to run any you know I think what happened why frisky and and and other uh warnings went unheeded is that we all knew about these things but intuitively thought we could put a lid on them that they wouldn't be that big a deal and it turned out human intuition just isn't equipped to know how big a deal it was and that a little bit of data snooping could lead to an awful lot of error uh in a in a beautiful one of those beautiful coincidences uh that some people would would suggest was there was some meaning behind it in a beautiful bit of serendipity as I was driving home before we started recording I heard you speaking on BBC Radio um about AI I wanted to to to maybe ask you a couple of questions about Ai and data um towards the end of I think it's the second chapter of how the mind works you say that you you make the not the promise but you suggest that we by the end of the book you'll answer the question that's implicit in the title and then by the end of the book you conclude that we maybe we lacked the cognitive architecture to understand the human mind because that is that a fair characterization I wouldn't put it that way um I think there's certain questions that the human mind can pose that it may be and keep capable of answering They Don't Really pertain to how the human mind works in the sense of what makes intelligence possible how do we recognize faces understand sentences make coherent plans uh I I think we could have you know in principle a complete understanding of that but when all that is said and done and that's a lot that's really what Psychology and Neuroscience are all about there's still what you might you can call philosophical problems I hate to use those words because I'm married to a philosopher and she hates it and I think a lot of philosophers do when you use philosophical to talk about things that are unsolvable Airy fairy metaphysical uh but still it captures the idea something like this would be concrete how do I know that when you're seeing something that's read and I'm seeing something that's read what you're seeing is the same red that I'm seeing of course you call it red but maybe if I could be inside your head I'd call it green and I've just been using the word red I prefer the green things all my life you know that's an example or when I make a choice uh is it was it all kind of predetermined at the point of the Big Bang because it's all a bunch of physical deterministic processes including the ones in my brain or can I sometimes somehow transcend cause and effect and choose something independent of everything that's happened before or is right and wrong uh really out there in reality or is it just a construction of my own brain questions like like that uh might be ones that we kind of spin around and around in circles just because they're we can ask the question but it may not be ansible not because you know and some folks some philosophers would say well that just shows that they're meaningless questions if they're not answerable principle then what meaning could they have I wouldn't go and I wouldn't take that step that logical positive step I think they're meaningful but they still might be unanswerable but one way of thinking about it is that it's almost inevitable those types of problems could occur if you have a scientific mechanistic understanding of the Mind the mind is a gadget it's not we're not Angels we're not there's no guarantee that our mind can think any thought conceive any explanation maybe for the same reason that chimpanzees can't understand prime numbers that doesn't say anything about prime numbers it's just something about chimpanzees maybe our own brains make it hard for us to think certain thoughts anyway that is an argument that's been put forward by a number of philosophers Colin mcginn's probably the most famous contemporary one I think there may be something to it but even if true it leaves so much work to be done and so much genuine insight to be had on how intelligence is possible which I don't think is mysterious that there's lots of satisfying work to be done and understood do you I'm interested to know what you what you think about the the Apparently sudden appearance of this new generation of AI tools and platforms from the the very much uh convergent rather than Divergent thinking chat GPT to Mid Journey um do they excite you do they terrify you do they make you think we may um have a Sorcerer's Apprentice that can help us to answer some of these big thorny problems um I I don't think they will help us to answer the thought any problems like how does brain activity result in subjective experience the one thing they themselves are dependent on the output of the human mind beforehand namely all that stuff that we've dumped onto the internet that it's trained itself on um so I I I doubt it um I'm at this point still puzzled and I have to confess part of my possiblement is that I'm dissatisfied with the explanations that are out there on how they work and uh when I have some time I want to take a deep dive into the kind of the guts of these programs just to answer the question how are they how are they accomplishing something that most of us would have said a system like that couldn't accomplish that is and I I confess to being surprised I think I'm not the only one that they could put forth such coherent answers on pretty abstract uh tasks paraphrasing Pros summarizing answering questions in a coherent way partly because perhaps this goes back to ibusters law of small numbers that uh perhaps we're we're just not equipped to anticipate in our own mental simulations What A system that process a petabyte of data and summarizes its regularities in a trillion parameters what is a system like that capable of more something more than I thought what's the trick where in those trillion connection weights are is there an understanding of the gist of a sentence or a logical relationship between all and some it's kind of buried in there somewhere how did it how exactly did it get there how does it get out you know I confess that my understanding is so too superficial to be able to answer those questions and until we do there's a lot that we we don't know it's kind of a almost an alien intelligence that we've got to study the way we study our own intelligence it's not like ours in in many ways but it is uh it does impressive things and and fascinating I mean you're going obviously completely right about you know it depends on what we've dumped on the internet and and and then it then generates its own stuff that it uses in ways that it it understands how how it uses them one of the things I found it to be particularly poor at doing in many different platforms is jokes um really bad really hopeless at puns really bad at that brilliant comic thing of the switcheroo it seems to have even though it's kind of you know it's got all of the it's got all of the the Corpus of of 20 years 20th and 21st century humor um it's a draw on uh it seems to be singly bad at doing that that brilliant I mean I often think that that um that the particularly the switcheroo when you're a comedian is taking you down uh one line and then completely flips 180 and that's that's the cause of the humor um I I often think that's that's kind of akin to to the human faculty of insight I quite like to come back and talk about about inside just thinking on a on a uh one other question on AI um you will know Harare writing The Economist recently added his voice to the growing list of thinkers and academics and policy makers and regulators and also many of the investors in open AI two um calling for a moratorium on this Mass rollout he he's talking about the creation of a regulatory Authority like the Food and Drug Administration perhaps um to make sure these tools are safe what do you think do you think that's necessary or or do you think the the danger that they pose is greatly exaggerated I mean I think there are dangers and the obvious one is industrial scale misinformation um the uh and and potential for fraud since they can impersonate a person pretty well do fake videos uh we don't know what the counter measures will be because people don't like to be lied to at least you know a lot of us don't like to be lied to and so we're not going to just uh swallow everything that can be thrown at us we don't know how effective they kind of the the Paper Trails or the fact checking mechanisms are going to be to know whether you're seeing reality or a deep fake uh but they're going to have to happen because uh I don't want to watch a bunch it takes to be told it's news so maybe I'll rely more on you know the BBC in the New York Times and uh CNN as opposed to some random uh uh social media feat uh and I think as with anything if there is deliberate harm or exploitation it's perfectly legitimate for there to be laws to to prevent it I don't I think moratorium is just too blunt and coarse uh it's you lift the moratorium and you're right back where you started uh you I think we have to know what the specific arms are by seeing what will be what needs regulation in the sense that ordinary countermeasures spontaneously arising aren't sufficient uh this is a fraud we already have laws against fraud maybe they have to be beefed up when it comes to AI generated fraud uh in cases of uh some of the threats of uh control of infrastructure well I think just as we don't turn over any infrastructure like the electrical grid to any old amateur we've got to have safeguards that untested systems aren't given uh control control over a an airplane whether government is necessary and at what level whether it's opening up um avenues for lawsuits for for damage for torts where there has to be a priori regulation I think it's going to have to be done on a case-by-case basis uh that is we have to know what the harms are what harms are properly disincentivized by law and Regulation and uh sweeping across the board uh throttling is doesn't seem to me the most the most productive uh Safeguard could I could I can I ask you about a little bit about about Insight how you get how you move from data to inside here's a here's a another one of these sort of metacognitive problems you know what what is inside um perhaps you've got a much better definition than me I I like to think of it as being this kind of profound and useful understanding it's profound because it answers the not the trivial but the the entry-level question of what do the data mean so what and allows us to go on and do rather more interesting things and say now what what shall we do as a result um if you have a much better definition please tell me um but what I'd like to know is how you go how you get to to inside how you get in your papers in your books to move from we've looked at this this is what we they seem to be showing and therefore this is what we conclude as a result yeah so I I I agree with you definition of insight and as a cognitive psychologist the puzzle is things how does it actually work what's going on in the in the brain that that gives us these Strokes of insight as I you know I don't like to believe in magic I don't like to believe that we've got some kind of spirit or soul that has an uncanny ability to to to to to see true things I mean that's just too mystical so something is happening that's mechanistic I think they're um and there's been research in classic artificial intelligence uh as well as massive neural networks uh that address that one source it may not be the only one but I think it's a very important one is analogy um and this is one that I I talk about in my book the stuff of thought and here I rely on in part on the work of the cognitive psychologist edry gintner um and and philosophers like um uh uh Robert Boyd the um an awful lot of scientific insights came from analogies um not superficial metaphors this reminds you of that but rather that the kind of the flowchart the logic diagram the Deep underlying laws can be ported from some domain that we're familiar with to some new domain that we're grasping we're grappling to understand so you know the uh some of the obvious examples are the solar system is a model for the atom in the work of Niels boy um the uh a code a linguistic code as a model for the heredity in Watson and critch um the uh flow of water as a down a waterfall as a analogy for the transfer of heat in thermodynamics um billiard balls in in terms of um statistical mechanics and thermodynamics that heat is a bunch of billiard balls bouncing faster faster the thing about that and the what makes it a challenge is you can't just think about any old something it reminds you of because there's a lot of pseudoscience that relies on uh superficial analogy like powdered rhinoceros horn is a cure for erectile dysfunction because you know it's you know long and pointy and something else is long and pointy therefore one treats the other you know that's not the way to do it and there's nothing in particular that is perceptually similar between a you know say a telegram and a stretch of DNA or an atom in a solar system is rather the underlying forces and interactions in the case of uh say the solar system it's centripetal force and attraction and momentum in the case of DNA and say a telegram it's redundancy and information content and um stop and start signals he said abstract level in which you could actually say um not just that two things are similar that they remind you of one another although that's probably the start but that at some level of abstraction they're actually the same that is you could talk about centripetal systems and atom being an example a solar system being an example a tetherball being an example and that at that level of abstraction they really work the same way the explanation is the same and the reason that this is I find appealing or in the case of say uh selection there's natural selection there's artificial selection what you know pigeon fanciers do there may be selection among uh ideas there's um uh artificial evolution in computer systems but that in all cases you can talk about random variations selection multiple generations of copying they really are the same at some deep level the reason that I I find this a promising way of thinking about insight and and the small conversation really does get back to your question I promise uh because the thing about Insight is it can't the puzzle that we have to solve the reason that I think you brought it up is that Insight can't consist of any old random Association you know this makes me think of that well big deal anything can make you think about anything the The Challenge and understanding Insight is why the idea is good ideas any ideas good ideas why when you go Eureka or a light bulb goes off doesn't really solve the problem as opposed to Just Lead You In some random chain of associations if there are I promise to wrap up if there are these underlying systems like a centripetal system uh selection by consequences information codes that underlie reality in different superficial manifestations then one of them really can give you the answer about another give you insight into the other assuming they both legitimately are two examples of the same underlying purpose system so totally totally with you on analogy and I think that one of the one of the the traps that people very very often fall into in data storytelling as I would as I would term it is or one of the ways out of the traps is when you is when when a simple analogy is used to to to to quantify or to contextualize um uh data I think analogy can be incredibly powerful um I mentioned that I'm a big fan of the sense of style um and you do you do you devote a whole chapter to the curse of knowledge which I think is is and you point the finger at uh as academics and financial advisors and lawyers and the rest of it um for our listeners not familiar with the phenomenon Could you um describe it and could you maybe talk about um the connection between the curse of knowledge and when people maybe particularly act some academics use too much data yes right so the curse of knowledge is a psychological phenomenon uh in which uh once you know something it's extraordinarily difficult to imagine what it's like not to Know It uh that is to uh project yourself into the minds of other people who don't know what you know even to remember back to before you knew it uh people think that if you know it now you always knew it and the classic demonstration comes from an experiment with children in the case of um uh cognitive development the same phenomenon is sometimes called uh lack of a theory of mind or mentalizing the theory here not referring to a scientific theory but to kind of our own everyone's folk Theory so the classic study is you take a three-year-old you play a little a little scenario um you take a marble that was in a basket and you put it in the box and the child sees it they know John knows what the marble is and you say um well uh sorry Sally see Sally a puppet sees the marble in the basket Sally leaves the room you put the marble from the basket into the box Sally comes back into the room where will Sally look for the marble and kids will say in the box now the child knows that the marble was put in the box but Sally wasn't in the room when that happened Sally has no way of knowing uh that the marble was in the Box the child just instinctively assumes that what he himself knows everyone else knows and gives the wrong answer about Sally and by four years old four years of age most kids um solve the problem they know that what they know and what Sally knows are different still it's uh it still afflicts us as adults and I I think the main explanation for why academic writing is so Dreadful is not the common conspiracy theory academics they try to sound highfalutin they're trying to bamboozle people because they have nothing to say with all this highfalute jargon uh to make up for the fact that there's spider banalities I actually don't think that's the the best explanation because I know a lot of brilliant Scholars who um really do have something to say but they just can't say it and the reason they can't say it is they don't know that their entire autobiography you know umpteen years of graduate school and postdoc and being a professor and they've a mass all of this technical vocabulary all of these paradigms and analytic techniques each one of which they can identify with a label and they think that everyone knows those labels and are they doesn't occur to them that Their audience hasn't lived the lives that they have lived and a lot of it is completely needless so in in you know in neuroscience uh there's absolutely no Precision to be gained in talking about a murai model as opposed to a mouse um and or for something like in in plant biology arabidopsis if you just say you know in parentheses a mustard plant you know you've sacrificed you know a dozen keystrokes and you know that amount of screen space and you've multiplied by several orders of magnitude the size of the audience that knows what you're talking about so it's a bad trade-off to use abbreviations and or to fail to give examples when doing so multiplies or maybe even exponentiates your your readership and also it's not just abbreviations because those are the easiest things to weed out um but also abstractions uh describing something in language of what it means to you as opposed to the concrete goings-on so if I I'm gonna you know lose my own field of cognitive development or one of my Fields you say an age-appropriate uh engaging stimulus uh was used in uh condition one uh if you just said a Mickey Mouse doll uh then people would be able to form an image and know what you're talking about age-appropriate you know developmentally appropriate stimulus what does that even mean uh and there's a lot of scientific and academic talk which is couch in abstractions and the human mind is best when I can visualize something and and it takes us I'm right outside that it takes us clearly it takes us longer to process and work out what that means in exactly the same way as uh I mean maybe this is a a particular example but if we use the if we use the passive voice rather than the active voice it takes us longer to process what that is the the cat chase the mouse is easier and quicker to process than the mouse was chased by the cat yeah um in general although I have a discussion of that and that's a kind of advice that is age-old it isn't exactly right there are actually a lot of cases in which the passive is better because and and kind of appreciating how language works the logic of language can help you differentiate between when the passive actually might be the better choice so here's the thing about language uh and it is a a question of information data if you want language has to do two things at once or at least sorry the order of words in a sentence at least for languages like English which rely a lot on word order so you've got your word word word word how do you know who did what to whom well in general the word before the verb is tends to be the doer and the word after the verb tends to be the done to so we use the left to right position or early to late position in a sentence as a way of coding information about who did what to whom that's fine as far as it goes except for the fact that were Mortals were serial information processors so words that come in earlier in the sentence are psychologically different than words that come in later in the sentence that is the first words Orient you it's how you sort of set a scene where you know what's being talked about the later words kind of comment they add information to the context that you've set up the scene that you've set well that means that there's contradictory demands on word order word order is being asked to do two things at once on the one hand to code information on the other hand to engage The Listener or reader's attention the reason that we have the passive voice the reason it's been in the English language for you know more than a thousand years is it's a way of decoupling them if you can say the mouse was chased by the cat you're expressing the same content who did what to who but you're getting the mouse out earlier in the sentence and if the entire preceding context was about the mouse if so to speak to use a visual analogy the camera has been trained on the mouse in all of the sentences so far the metaphorical camera and now a cat peers out of the blue and Chomps the mouse it's actually easier to understand the sentence if you say the mouse was bitten by the cat because the mouse is how you come to the scene in the first place based on a prior context so in that case if you've been following the travails of the mouse for sentence after sentence after sentence if you then say the cat ate the mouse it's like oh wait a second what cat uh if you say well the mouse was eaten by the cat then you have that continuity and coherence so as a modification of your advice it's not that the active is always better than the passive the passive is better when the attention has been focused on the done to the the patient to use the technical term but if you pardon an example of itself the passive tends to be overused in academic uh and that is a that is a genuine problem that it that it is overused because um the the problem there is that the writer again it's a kind of cursive knowledge problem the writer kind of knows what's happened they start with the outcome because they know the outcome and then they kind of work backwards and tell you how it got there whereas the poor reader doesn't know how it ends up they want to be led through a narrative um you know cause before effect the tendency to have effect then cause is an example of the cursive knowledge or someone who already knows how the story ended up and that's why I can one of the reasons academics overuse it that is that you you may have you may have seen um the appearance of my own cat Quincy's tale during the previous answer he was chasing I have to say no mice um I have a Colombo question for you Steve just before we close um in the way that you work with data is there anything that I haven't asked you that I should have asked you I guess you know lots of potential things you know and things that I'm sure should turn up in your show all the time like uh how do you know when to trust data data can be you know you can you can you can be led to wrong conclusions with confidence if you take the data at face value especially correlational data uh what are are the kind of this Cutting Edge state of the art techniques for uh Discerning causation and correlational data uh and you know here I gotta confess I'm in the professor's situation that my students who later generations are closer to The Cutting Edge than I am um how do you know whether um you know when a trend can be extrapolated or when when it's say a decline in war is this just a um could might we be fooled by a random pause at a fundamentally stochastic phenomenon that the fact that they say hasn't been a major uh great power War for 70 years may not indicate a change in the underlying probability of the generating process as opposed to the inevitable gaps in any sputtering process not a trivial a question well when I wrote the better angels of our nature I I you didn't really have the means to answer conclusively since then much more sophisticated data scientists than I have tried to dissociate random pauses from changes in the underlying generator and have satisfyingly Vindicated my hypothesis that there is an underlying change but it's a real challenge in data science to tell them apart whenever you have a stochastic process finally now it is not remotely difficult to find an awful lot of about use uh Steve online where would um where would our listeners our viewers find what you what for you is the most authentic or the or the best curated place to find about find out about you and your work added maybe your photos online okay uh well in in my own website stephenpinker.com uh there are my courses including my course lectures my articles Pages for each one of my books um a link to my photo site which confusingly is called stevepicker.com but linked is there uh together with a page I call Silly uh with various caricatures satires funny little uh episodes that are featured on the web fantastic we'll link we'll include a link to that obviously in the show notes look thank you so much for your time Steve thank you for sharing your approach to the way that you over many many years in illustrious career have used data in a smarter way I think if there were only more people with your pragmatic approach there'd be rather less data Malarkey in this world and rather more data Common Sense Stephen Pinker thank you so much for joining us today thank you sir thanks so much for listening to data Malarkey the podcast about using data smarter to find out what kind of data Storyteller you are why not take our data storytelling scorecard it takes just two minutes to complete and will give you a personalized report right away visit data storytelling.scoreapp.com or follow the link in the show notes foreign

Channel: Data Malarkey

Views: 15,087

Keywords: data malarkey podcast, sam knowles, data storytelling, telling stories with data, data, data analysis, data analyst, podcast, data science, storytelling with data, data analytics, data malarkey, how to tell stories with data, data driven marketing, digital marketing, data communication, data story, data scientist, marketing, marketing data, steven pinker, common mistakes of data storytelling, steven pinker interview, steven pinker harvard, steve pinker

Id: 0x65odw-SvA

Length: 48min 41sec (2921 seconds)

Published: Wed Jun 28 2023