By Glitch Team

November 17, 2016

Rethinking Intelligence

In this Tech Talk, Nancy Hawa, Support Team Lead here at Fog Creek, explains why we need to rethink our attitudes towards intelligence. She puts forward a broader, more inclusive theory of intelligence than the one typically held by many in Tech. We’re encouraged to think of intelligence more like a skill, and we learn why it’s important to do so. Finally, Hawa shows us how our biases and judgments of others can impact outcomes and what we can do about it.

At Fog Creek, we have weekly Tech Talks from our own staff and invited guests. These are short, informal presentations on something of interest to those involved in software development. We try to share these with you whenever we can.

Contents #

Transcript #

What follows is a lightly edited transcript of the talk:

I developed a version of this talk while I was a student at Flatiron School. It was substantially different from what I’m talking about now, but the reason I gave that talk was that I noticed that many of my classmates were feeling really, really frustrated. We did this thing called ‘Feelings Friday’, and everybody would talk on Friday about how they were feeling. At some point, the instructor said that he was going to go into hard mode the next week because it seemed like we were doing well, and one of the girls started openly sobbing. People were feeling frustrated, and they were feeling dumb, and it’s because, I think, the tech community, or when you start getting into coding, there’s this almost cult-like worship of intelligence. It’s a very limited idea of what intelligence is, and I think it can be really bad for most people trying to come in from the outside, and also bad for us if we’re already on the inside. I’m going to talk about why that is.

So, what do we already believe about intelligence? I think everybody has seen something like this. You’ve all heard the term IQ score, and we have people take tests, and then we put them somewhere on this distribution and most of the people are in the middle, smart people, are over on the right, and the dummies are on the left. This is a very simple, one-dimensional idea of intelligence. You are smart or not so smart, or somewhere in between.

I think that, intuitively, we also have an idea of intelligence that looks something like this — where there are people who have word smarts or math smarts in varying degrees. Maybe you’re someone like this who has a lot of math smarts, but you’re not so good with the words. Or maybe you have a lot of word smarts, but you’re terrible at math. Or maybe you’re just bad at everything. I think that people map onto this partly because they make us take the SATs in high school. We get scored on both of these things.

Then, if you’re me, you go to education school, and they show you another diagram. One that says there are eight kinds of intelligence, and they are all equally valuable, and all equally meaningful. If you’re like me, the first time I saw this, you look at this, and you think this is bullshit because word smart and math smart are the real smarts. Probably some of you are thinking, really, it’s just logic smarts and nothing else matters at all. That’s not an abnormal thing to think, but it’s not a valuable thing to think.

I have come to believe that all of these things are equally valuable. I was initially like, “nature smart”? that is ridiculous. That is not a kind of intelligence. But, the way we talk about these different kinds of intelligence just casually can be very devaluing. We call people who are good at logic ‘smart’. We call people who are good at music or art ‘talented’, and that word talented is almost disparaging in that context. It’s like saying there’s smart, and then if you’re good at music, that’s not because you’re smart. That’s something else. We’re not going to give you the word smart. I think that that’s bad. The important thing to take away is that any of these things are things that you can develop. If you have any of these things, you can use it to develop any of the others. There’s a lot of evidence that you can get better at math from music, or better at music from math.

If we think about all of these things not even as kinds of intelligence, but just as skills that you can develop independently, we get to a completely different place.

Intelligence is immutable, but skills can be learned. We think that people are as smart as they can ever be. People who take the IQ test at three or five, the idea is that we’ve determined how smart they are, and that’s how smart they’re going to stay. But you can learn new things. Judging people by intelligence early, and putting a label on them and saying you are or are not smart, is damaging because it makes them feel as if the learning part is not the main thing, and the learning thing is.

The bigger thing of what I’m going to talk about more is that “smartness” is very culturally charged. The very idea of ‘smartness’ is culturally charged. How smart you are, according to conventional standards, is also affected by culture. So, surprise, this is also a talk about diversity.

Greenwald and Benaji developed a method of looking at implicit attitudes that has been used a lot since. The problem that they were trying to solve is, if you want to guess what someone’s attitudes about something are, about race or about gender, it’s hard to figure out because you can’t ask them. Most people don’t say that they’re racist or sexist, and a lot of people don’t even know. A lot of people would honestly and truly answer I am not racist, I do not have any biases about race at all, and they are wrong, and they do have attitudes that are about this. So they wanted to come up with a way to measure peoples’ biases.

What they did is, and this is a computer task, and you sit there, and I am using one about gender and science. There’s a lot of different versions of this. They say, all right, put your finger on the E and the I key. If you want to move this word ‘boy’ to male, you press I, if you want to move it to female, you press E. Then you go through and you see a list of words like boy, grandmother, grandfather, father, and mother, and you have to identify if they are male or female. So this one is male. They time how long that takes you to do it.

Then they do the same thing, but with liberal arts or sciences, and I have to put literature toward liberal arts and chemistry toward science. Then they give me both at once, and they say you’re going to get chemistry and grandmother, and you have to put chemistry and father over here, and sociology and mother over here. Then they switch it, and female goes with science, and male goes with liberal arts, and they are timing me again, and the difference between how long it takes me to do the task where male is with liberal arts and female is with science, and how long it takes me to do the other one shows something about me. If it’s really hard for me to do this association test with men on the liberal arts side and female on the science side, that shows that I have male/science, female/liberal arts association.

They also do this with race where they show black and white faces and good and bad words, and you have to send black faces to the good side, and white faces to the bad side. The first time I was asked to do this, I was in college and I took the test, and they tell you at the end how racist are you. I can tell you that I am certified not racist by this test, but I also can tell you that I felt the challenge of putting black faces on the same side as good words. Like, a black face goes with the word terrific and a white face goes with the word terrible. I hope you won’t all judge me for that.

What we find is that there are really strong links in people who say they are not racist, like me (I say I’m not a racist), and people who believe they are not racist, to people who believe that they don’t have this implicit association between gender and scientific-mindedness, and people are wrong. They don’t have this kind of study specifically for words of intelligence, where you would sort it into smart and stupid. But I’m going to take a leap and say that I think that people do have those associations. People who have trouble sorting white faces to bad words and black faces to good words, many people are going to have these kinds of implicit associations, and it’s something that goes across races and genders and social statuses. People are really surprised by what they find.

This means that it’s sort of unavoidable that you’re going to make snap judgments about how smart someone is when you meet them. Or how good at science a woman is going to be when you meet her. This is really dangerous. The fact that looking at someone is going to be enough for you to color your opinion of how smart they are because of something called confirmation bias.

Basically what confirmation bias says is that if you believe something, when you get evidence, you’re going to listen more to evidence that confirms your belief than you will evidence that goes against your belief.

This is another study that they bring people in and say, are you a proponent or an opponent of the death penalty, and they write down how everybody stands. Then they say, okay, we’re going to have you read two studies about the death penalty, and we’re going to see how you feel about it afterwards.

They give them a card with a one sentence summary on one side of it to see if that one sentence summary affects their beliefs. It turns out that that one sentence summary doesn’t really affect anything because they already have the belief before. If people who start out as proponents of the death penalty, the study showed how much people’s views changed as to whether they become more in favor of the death penalty or less. People, after the pro-deterrence one, proponents of the death penalty are changed — they’re even more pro-death penalty. After reading the anti-deterrence one, they’re a little bit less in favor if they just see the results. They don’t read the study, they just see the results. But seeing the results of both, they get both studies, one says one thing, one says the other, and they’re fake, they’re designed to be about equivalent. But if participants got the evidence on both sides, they were even more on the side that they already believed. So, if you get evidence on both sides, the same evidence, you just believe what you already believed more.

So if you believe something and are presented with neutral information about it, and these people are presented with the same two articles, if you were opposed to the death penalty, you are now more opposed to the death penalty. If you were in favor of the death penalty, you are now more in favor of the death penalty from reading the same things. If you don’t just get the results, you get all of it, you just get more and more and more believing what you already believed. Even if you read the thing on the other side, that still makes you more sure that you were right if you have the whole thing and not just the results.

Why that matters, why confirmation bias matters, is that we also know that people have confirmation bias about their opinions of people they meet. So if you meet someone, and your initial impression of them is that they are not smart, perhaps because of a bias that you have about their race or their gender or what they look like in some way, you will find evidence to confirm your belief. We know that this is a human way of thinking. We want to find evidence that confirms our own opinion.

However, there’s research that shows that one way to limit the effect of confirmation bias, is to tell people that confirmation bias is a thing, explain it to them, and have them thinking about it. So it’s possible that jus by reading this, that I have just fixed you

But, this is something that we should be thinking about because if you get an opinion in your head that someone is not smart, that’s something that you think cannot be changed. Whereas if you’re thinking about whether or not they have the skills that is something that’s not immutable in your mind in the way that intelligence is.

In this study with students at Princeton, and one limitation of the psychological research is that 90% of it is conducted on 18–22-year-olds. So in this study of Princeton students, there were 40 black students and 40 white students. They’re asked to perform a task that’s related to golf, where the better you do, the fewer strokes you have is better. So low scores are good.

There were four conditions. There’s the natural athletic ability condition. These students are told that this is a test of natural athletic ability. The athletic intelligence group is told that this is a test of athletic intelligence. Both the race prime and the control group are told that it is a general test of athletic performance, but the race prime group is asked before the task what their race is, and the control group and all of the other groups are not asked about their race until after they do this golf. They measure two things. One is how many strokes does it take you to do whatever this golf thing is, and the other is before the golf thing and after the golf thing, they do a five question measure of your anxiety level.

The results on the number of strokes show that with white people if they’re told that it’s natural athletic ability, it takes them 27.8 strokes. But, if they’re told that it’s athletic intelligence, they do much better. They do much better if they’re told that it’s a test of intelligence. Black people do much worse if they’re told that it’s a test of intelligence. So saying to a black person, this is a test of your intelligence, makes them more likely to perform poorly. Saying to a white person this is a test of your natural athletic ability makes them more likely to perform poorly. Asking the participant about their race before their start makes them do much worse if they’re black, and doesn’t have much of an effect if they are white.

Let’s look at anxiety. So the black people were much less anxious when they were told that it was a test of their natural athletic ability, but they felt much more anxiety when they were told that the test was about intelligence. Both levels of anxiety interestingly go down if you ask them about their race. And we know from many other studies that this effect is replicable. They also did a study at MIT, for example, with Asian women and divided them into three groups. One group was asked about their gender. One group was asked about their race. And one group was asked about neither. The women who were asked about their gender taking a math test did worse. The women who were asked about their race did better. So we know that priming people and talking to them about their race has an effect on their performance. We know that the anxiety level effect of stereotype threat is real, it’s duplicated across multiple studies.

The Pacific Power and Lighting Company, a power company that had power lines going through the woods, over the Cascades Mountains. Every spring they’d have lots of ice storms that created a large accumulation of ice on the lines, which you can imagine would be damaging. The towers can fall, the lines can break. The method used to remove the ice was to send linemen into the field, have them climb these poles, and then they had a pole with a hook at the end, and they had to climb to the top and then shake the lines off to get the ice off. It was dangerous and really, really unpleasant.

Everybody involved agreed this is a problem that we want to solve. They did a lot of brainstorming sessions with no results, and then they brought somebody else in. They could have defined this as an engineering problem and said we’re going to just talk to our engineers about it. They could have said we’re just going to talk to the linemen and their supervisors and have them hash it out, but this consultant guy that they brought in said, no, everybody in the company will be involved. It’s going to be accountants, secretaries, people from the mail room, everybody.

They didn’t get anything out of it for a little while, and then they’re on a coffee break and two of the linemen are chatting, and one of them is telling a story that he comes down from one of these lines, and he sees a bear and the bear chases him. Somebody else says, hey that’s an idea. Let’s just train the bears to climb the poles, because they’re so big and heavy that their weight would knock the ice off. So this is not a productive idea, they’re not going to train the bears. But, they’re joking and laughing, and somebody else says, well we don’t need to train the bears. Let’s just put honey-pots on top of every pole because then the bears will go up on their own. Again, not a great idea. But this isn’t a serious brainstorming time. People feel free to just say anything. Somebody else says all of those fat cats in the main office, they fly around in helicopters. You should just use those helicopters to put the honey pots on the top of the poles, and then all those fat guys will have to get some exercise. Still not a great idea. But then someone speaks up and says that they were a nurse in Vietnam, and they would bring the injured soldiers from the helicopter. There was so much down lash from the helicopter with dust flying, I bet that if you just flew the helicopters near the power lines, that would shake the ice off. And that was the solution.

The solution to what could have been defined purely as an engineering problem came because they put a bunch of people who were different from each other in the same room, talking about the same problem.

Your perception about someone else’s intelligence may not have very much at all with what you’re trying to measure. Prioritizing intelligence can be exclusionary, and that exclusion does matter. Diversity does matter.