Monday, September 16, 2013

Ignoring the negative

by Matt McCormick

“Isn’t it weird how celebrity deaths always come in threes?”
“I’m telling you, Asians are bad drivers.” 
“I know it’s not politically correct, but it's true, women just aren’t good at science.”
“I swear I have special dreams.  I dreamt the night before that my mom was going to have a car wreck and she did.”

Confirmation Bias is the mistake of selecting evidence that corroborates a pet hypothesis while ignoring or neglecting evidence that would disprove it. It’s the mother of all fallacies.  And the reason it persists is that it feels so right.  When you’re making the mistake, the conclusion you’re drawing has that shiny, aura of truthiness to it.

Humans are guilty of committing it in a wide range of circumstances. At the end of each semester, many students, including students in my (Prof. McCormick’s) Critical Thinking and Theory of Knowledge where we study confirmation bias extensively, blunder into it. They get a grade for the course that is surprisingly low and send an email to their professor asking to know what happened. As far as they knew, they were doing great in the course. They recall getting an A on an assignment, and doing pretty well on the midterm, and feeling pretty optimistic, so they can’t understand the low grade. Here’s a couple of real emails: 

Student Email 1: I just checked my grades for the Spring semester and was surprised to have earned an F. I completed the major assignments for the course and did well on the midterm (90%) and well on the final (85%). I know I didn't participate in the online forum as much as was required but I'm still confused about the grade. I took the class material seriously and did my best on every assignment assigned.

Professor McCormick (notice the grades highlighted in red here): Here are the grades I have for you. This syllabus gives the details about the grade structure. Check the math and check your returned assignments to make sure it's all right. If there's a clerical error, I'll fix it right away:

Question Sets: 0, 82, 0, 75, 0, 95 (6% each)
First paper: 78
Midterm: 90.5
Second paper: 85
Final: 85
Outside projects: 7/8
Google Group: 0/8
Attendance and participation: 0/8

So between the skipped question sets, the Group discussion and attendance, you gave up 34% of the grade. Even if you were making an A on everything else, that would put it down to a D.

Student 2: I'm emailing you in regards to my final grades. I was hoping you could provide a detailed summary of my grades for the semester so that I can understand how I received a D. I felt as though I did fairly well, particularly improving on the more major assignments, so I would just like to know how I still failed to pass. If you could email me a detailed summary of my grades, I would greatly appreciate it. Thank you.

Professor McCormick: Yeah, I was disappointed in your grades too. It seemed to me that you are capable of doing much better work, and being more responsible about turning stuff in. Here are the grades I have. Check the math with the grade structure on the syllabus and let me know if there is a clerical error asap: 

Question sets: 76, 0, 82, 0, 85, 56, 80 (6% each)
Evil paper 65
Midterm 68.5
Second paper: 75
Final exam: 85
Outside Projects: 6/8
Google Groups: 4/8
Attendance and participation: 6/8

So the skipped question sets took 12% off of your grade. You got a D on the first paper and didn't take the opportunity to rewrite it that I gave the class. You could have brought that up substantially. The Google Group points would have helped too since your overall score came out at 68%.

When we commit confirmation bias, we cherry pick the evidence that suits us. The student actively remembers the good grades, but missed assignments and low scores are forgotten.  Someone picks out the bad Asian driver, or the woman who does poorly at science, and then uses that to fortify their mistake. 

One more example:  over 50% of people think they’ve had prescient dreams or premonitions.  So suppose that you have 20 dreams a night, 365 days a year, for 10 years.  That’s 73,000 dreams. 
Which ones are notable and remembered?  The ones that seemed to have something to do with what happened the next day.  The dream you had that seemed to anticipate your mother’s car wreck leaps out in your memory as an extraordinary coincidence.  In China, there’s a saying, “No coincidence, no story.”  But more importantly, there are 72,998 dreams that weren’t special or notable. 

Clearly, having an accurate and objective grasp of the relevant evidence would serve us well. We don’t want to ignore evidence indicating something negative, disastrous, or dangerous because it doesn’t suit what we want to be true. Imagine if a doctor acquired a skewed view of the evidence concerning a potentially fatal disease this way. Suppose the Secretary of State ignored significant negative indicators in the behavior of an aggressive and hostile foreign country. Suppose a potential employer asked you how you did in your Critical Thinking course in college, and then she checked your transcripts against your distorted memory. Suppose you spend thousands of dollars over the years on losing lottery tickets because the occasional wins stick out in your mind so prominently, while the loses are forgotten.  Suppose you spend time praying to God frequently, hundreds or thousands of times in your life, and on the rare occasion when something vaguely resembling what you prayed for came true, you count that as an answered prayer, while ignoring the thousands of misses.  

Further Reading

Matt McCormick
Department of Philosophy
Sacramento State

(Note: A version of this piece was published on Matt's own blog Atheism: Proving the Negative on 6.4.13)


  1. Matt, how are you inclined to account for the ubiquity of this phenomenon from an evolutionary perspective? Does dual process theory have anything to say about it?

    1. trong lòng nàng phút chốc bị khuấy lên, sợ hãi đến cùng cực, khiến cho nàng gần như phát điên, phải hét lên một tiếng chói tai.

      Trong mắt nàng, Vương Lâm đã không còn là một con người nữa mà đã biến thành một bóng ma, một Phệ Hồn ma đáng sợ!

      Gần như tđồng tâm
      game mu
      cho thuê phòng trọ
      cho thuê phòng trọ
      nhạc sàn cực mạnh
      tổng đài tư vấn luật miễn phí
      văn phòng luật
      số điện thoại tư vấn luật
      thành lập công ty
      heo bản năng, nữ tử nhanh chóng lui về phía sau, thâm chí một chút ý nghĩ phản kháng cũng không dám nảy lên. Trong đầu nàng chỉ có một ý niệm duy nhất đó chính là chạy trốn, trốn. Trốn!!

      Từ trước tới nay nàng chưa từng bao giờ hoảng sợ như bây giờ. Loại cảm giác đáng sợ này chẳng những huỷ diệt đi ý thức của nàng, khiến cho đạo tâm của nàng xuất hiện một vết rách thật sâu. Dù nàng có may mắn tránh thoát được một kiếp nạn này thì ngày sau tu vi cũng sẽ giảm mạnh, muốn đề cao cũng không chút khả năng nào.

      Thần sắc của Vương Lâm lạnh như băng, trong mắt là một mảnh âm hàn. Một bước, hắn đã đi tới trước người nữ tử đang bay nhanh lùi lại. Tốc độ của nữ tử này trong mắt Vương

  2. I agree, I think, with everything you say here, Matt. The phenomenon is ubiquitous. But so are, as more people are made aware of it through articles like this, claims that people one disagrees with are committing it. I wonder if at least some of these claims manifest a confirmation bias bias. That could be a problem, too.

  3. Thank you Matt. Well done. What might you say to the person who argues that there are advantages to Wishful Thinking; IE the Placebo effect?

  4. Thanks all. Randy, this is total speculation, of course, but it could be that if System 1 is responsible for the quick and dirty assessment of what's going on around Og the hominid, the error rate from committing confirmation bias is an acceptable trade off because of the advantages that finding confirmations fast and then acting quickly infers. So Og's cognitive system is prone to develop hypotheses too soon, on the basis of too little evidence, and then he finds confirmations of it more readily than counter evidence. If it's saber tooth tigers he's trying to find, or other threats, then it pays to have a high false positive rate. Better to start thinking there are a lot of saber tooth tigers around, and then whenever he sees something that even vaguely resembles one, run like hell. "Better to mistake a boulder for a bear, than a bear for a boulder." I have no idea how you'd empirical test this, or give a compelling evolutionary argument for how we got it.

    1. Interesting. On the other hand, if there really is a general bias to confirm, then we also tend to ignore the the evidence that something we've concluded is benign is actually a threat to our well-being. So it seems like there would have to be some sort of asymmetry in the environment with respect to harms and benefits in order for that to work out to our advantage. Maybe the Mercier-Sperber account of reasoning provides a more plausible answer: we evolved to reason cooperatively. We are not that good at disconfirming our own theories, but quite a bit better at disconfirming the theories of other people.

  5. Thanks Kyle. There are a couple of other well researched biases that seem relevant. There's a substantial psych literature on Motivated Reasoning, which is the disposition to view evidence or arguments that favor something you already believe more favorably, and to react to contrary or "preference inconsistent" information with excessive skepticism, and extra critical scrutiny. So when a politician of someone's beloved Republican party gets caught with an intern, he dismisses it as a minor transgression, but when a Democrat does the same thing, he screams for impeachment. I've got a blog post called "Defense Lawyers for Jesus," here: that was an epiphany for me.

    There's also some really cool research on Bias Bias, or the tendency to readily affirm when someone else is making some cognitive error, but not see yourself doing it. That seems relevant to your point.

  6. Randy L, I think Wishful Thinking, and the Placebo Effect, are common, but they are different from what I'm talking about here. And it's completely plausible that in some circumstances those cognitive errors could have conferred some advantages on early hominids. Placebo Effect is a positive or negative effect that people often have when they think they are being treated for some medical problem. It turns out that just believing that you are being treated has measurable effects. I think I'd probably lump Wishful Thinking in with Motivated Reasoning (see above). You really want to buy that red sports car, so you construct a rationalization whereby you deserve it and can afford it. Or someone's odds of surviving an illness are very low, but she builds up an account in her mind about how she's going to beat it. It might be that people who do this with medical problems have an elevated survival rate. I think the evidence in favor of that is more mixed than in the popular imagination. But suppose that positive thinking about your prognosis does improve your odds of surviving cancer from 10% to 12%. It's still highly likely that you won't make it. But that 2% is significant, and every little bit could help. So if it does improve outcomes, then you could easily see how a cognitive bias in this direction could have been selected for in evolutionary history. The down side, of course, is that we have millions of people wasting billions a year on lottery tickets and slot machines, and so on, because of this tendency.