The Whale of Human Bias

On today’s episode of Melissa Chases the White Whale in Long-Winded Fashion, we’re going to talk about human bias, the challenge of unlearning what you have learned, and the general difficulties of epistemology. Grab your spacesuits and hitch your belt to New Horizons: first stop, Pluto.

You’ll always be a planet to me.

The solar system has changed since I was a kid. We used to learn there were nine planets, and I bet a lot of us loved that weird little outlier called Pluto the best because we associated it with the lovable Disney dog instead of the probably more apt Greek god of the underworld. Remember the cranky-but-sassy villain from Hercules? Blue-haired dude called Hades? Same guy, mythologically speaking, but that didn’t stop us from dreaming about our little ninth planet and what lay beyond. And then in 2006, the International Astronomical Union shrank the solar system, leaving us with only a measly eight planets and a demoted (and not unique) but well-known dwarf planet. It’s science, nothing personal or spiritual or philosophical, but dear lord, you’d have thought the Giants had busted Eli Manning down to water boy, or the Pope had reclassified Jesus as a mere saint.

This isn’t religion, folks. It isn’t sports. We’re talking about science. SCIENCE! For goodness sake, this is supposed to be the realm of the impartial, the land of those led by the data and not by conviction. And yet…there are t-shirts and memes and memorials to Pluto’s planetary status, as if redefining it had smashed it to smithereens instead of merely drawing a lot of publicity to the generally challenging task of classifying celestial objects.

Speaking of the challenge of understanding celestial objects…

I had the immense pleasure of attending the Carl Sagan Prize lecture during WorldCon, which was given by Br. Guy Consolmagno of the Vatican Observatory. His talk was “Discarded Worlds: Astronomical Ideas That Were Almost Correct,” and I’d recommend it when you’ve got an hour. If you haven’t just now, the broad point is that science is habitually wrong. It’s part of the process of science, of course, but science is also capable of being wrong even if we have the tools to correctly collect or interpret the data. Why?

Human error, and human bias. Which is to say, even scientists, the gatekeepers who understand sound experimental design and the math that is important for making sense of experimental data, are capable (even prone to) calculation errors that go unchecked for really long periods of time, questionable interpretation of the data to fit their pet theories, and even, now and again, deliberate suppression of data.

…Which leads me to a little light statistics.

So there’s this thing in statistics called a p-value. Don’t ask me to get into the math, but the point of a p-value is to check whether or not a particular effect could have been arrived at by chance. It’s basically the Bechdel test of science, which is to say that it quickly sorts an experiment into one of two broad piles, but doesn’t actually say anything about whether the math can really be used to draw the conclusion the authors are trying to draw. It has become the golden standard of publishing, but it’s problematic. Fidgeting with the numbers to get a publishable p-value is enough of a problem that it has a name: p-hacking.

We’re not talking about scientists at the dawn of the Renaissance who are afraid of being burned alive for defying the church if they don’t doctor their numbers. We’re talking about modern scientists. I mean, it’s not like they’re at risk of anything important, right? Like their funding, or livelihood, or meaningful place in scientific history…

Ah. That’s the rub, isn’t it?

Let’s talk truth and consequences.

We live in a world where being wrong is the worst thing ever. Make a sarcastic remark that gets deliberately misconstrued as racist by some self-righteous internet troll? Your career could go down in flames. Decide that you disagree with some minor point of doctrine in your church? Heretic – get thee behind us, you are now shunned! Fail to get publishable p-values out of your experiments? Bad doctor, no tenure for you!

Fighting internet trolls is a hopeless battle. (Here’s a raised hand in favor of giving people a little more benefit of the doubt and a little gentler dissent when we think they’re wrong, though that’s an ongoing struggle for me.) Convincing some religious folks that it’s hot hurting anyone for others to disagree with their beliefs is a Sisyphean task on a good day. (And I get it, I do–faith can be a real house of cards, logically speaking, but if that house of cards is the only thing standing between you and the indifferent tornado of life, you’re gonna stick with it, even if it means the shingles of your belief are inflicting supersonic papercuts on the rest of the people who are trying to survive the storm, but that’s a story all on its own.) Persuading scientists that it’s acceptable to be wrong, however, should be as simple as not punishing them for doing science, i.e., being wrong in a lot of ways that help define the body of knowledge.

10 years later and academia has made a lovely about face to support researchers…

Imagine this shift in academia has happened. Imagine that being published isn’t about p-value or being right, but about communicating how things have or haven’t worked in order to help other researchers ask smarter questions. Science is doing the thing science is meant to do and sitting comfortable with the importance of being wrong. Hypothesis: we’re still going to be slogging against bias and unnoticed calculation errors. “What about peer review?” you say. Or, “What if we gave scientists genuine financial motivation to find basic calculation errors in the studies they review?”

All well and good, if you don’t mind throwing acid on an already frequently toxic academic dynamic, of course, but even if you’ve got academics taking each other down with the gleeful commitment of wizards vying to be head of the Unseen University, there are still limitations to the interaction of human bias and science and my reason for thinking this begins with socks.

Knit’s about to get real.

I’ve been knitting since 2007 and making up my own patterns nearly as long, though it took me maybe three years to start sharing patterns. By the time I designed and submitted my first sock design to a publication, I was pretty confident that I knew a thing or two about knitting.


So my pattern was accepted (yay!) and sent off to the technical editor (*trembles*) and then I get an email. “Hey, can you check your gauge again? Stockinette stitch shouldn’t be square…”

That’s weird, I thought. Stockinette almost always comes up square for me. I took some measurements of the sock in question and some close photos and sent back the evidence. My gauge was, indeed, square. It just wasn’t stockinette. For those of you who don’t knit, stockinette is essentially the most basic stitch when you’re working in the round. It’s the foundation of all knitting. It’s one stitch all the time and it’s the stitch that everything in knitting was based on and I WAS DOING IT WRONG. I was getting away with it because I know how to work with gauge and design modification (the math is pre-algebra stuff, we’re not talking rocket science here), but I was using a non-standard base stitch to get there in a way that effects both fit and wear over time…and is therefore a massive no-no in commercial clothing pattern design.

“Wait,” I hear you saying. “Isn’t this an example of peer review working to clue you in on something you didn’t know?”

Yes, yes it is, but there’s more. The tech editor, who is a gem of a human person, offered to Skype with me and help me figure out what I’m doing wrong. She took one look at my hands in action and said, “Oh, you’re wrapping the yarn wrong.” I frowned and fussed, “I don’t understand. A knitting instructor told me I was wrapping the yarn wrong when I was going the other way.”

She shook her head. “I wish people wouldn’t try to teach when they don’t know what they’re doing. You must have learned how to knit the way they do in Eastern Europe, so your stitches are oriented differently on the needle, which makes a difference. What you’ve got to understand is that what matters isn’t so much how you make the stitch as how the stitch comes out. You’ve got to understand the nature of the fabric.”

I’m paraphrasing, but the general point was a good one, and it’s pushed me into spending a lot of time thinking about how we arrive at a set a beliefs about what we know, and how we break them if we don’t have someone with a legitimately deeper understanding of subject X to look at our work and say, “You don’t even understand the question you’re trying to ask, do you?”

I swear, this post really is about science.

My friend Dan reminded me recently that science, while not perfect, is really the best tool that we have for not lying to ourselves. I agree with that statement. I also have a lot of confidence in at least the hard sciences to work through the bias bit by bit, because EVENTUALLY there’s bound to be a dataset in an experiment that points out the wrongness of some conclusions and we can make progress, even if that progress is tiny and incremental.

What is extremely, very not clear is how well the process of science can function when the parameters of study that make science work are not available for ethical or practical reasons. The psycholinguists in the room are probably chuckling creepily to themselves and imagining an experimental design inquiry into the importance of nature vs. nurture. I’m edging slightly further away from them and looking at my knitting and wondering how science can help me make sure that my next knitting design won’t be rejected for some other small but critical error.

And the answer hit me last night, with the help of my husband, the crappy wiring of our old house, and Doctor Who.

Let There Be Minor Electrical Fires

So John and I live in a house that has seen the turn of two centuries. Because the last century has seen an incredible leap in technology and availability of that technology for managing day-to-day tasks, our house has been put through several wiring changes and is due for another. We’ve been swapping out outlets here and there, but we still have a few that only have two prongs, which we compensate for by using the ground adapters, and like probably 99% of Americans with old wiring, we were using them wrong until I read that the reason they have the little metal tab that everyone breaks off is because if you take out the outlet cover screw and reinstall it through that tab, it will ground the outlet.

I shared this tidbit with John while he was in the middle of dealing with a possible wiring issue that was putting strain on his UPS, and he decided I was misinformed, which came out while we were discussing another old outlet.

“There’s no grounding wire to any of these outlets,” he said. “That screw just connects the tab through to the grounding wire screw. It won’t do any good if there’s no grounding wire.”

I squinted at him. “Why would an ungrounded outlet have a grounding wire screw?”

He opened the garage door and waggled a finger at me. “I’ll go get the old outlet so you can see!”

Thirty seconds later he comes in without the old outlet, but with a screwdriver and a grounding adapter that we have not broken the tab off. “Okay, so it doesn’t have a grounding screw,” he says. And sure enough, when he tested the circuit before and after attaching the screw, it went from “open ground” to “correct.”

I’ve been poking around the internet a bit, and it seems like there are plenty of circumstances under which this would not have worked, but our ancient outlets still use the metal boxes and conduits, so I lucked out and ended up being right to think John was wrong for blowing off the bit of advice I had picked up from some dubiously more knowledgeable source online. (So gratifying, that, and rare in the case of home improvement questions.) The only reasons I pushed the issue here, however, are (1) I have a history of being slightly less wrong than John where electricity is concerned, hey-o Radioshack! and (2) my baseline assumption is that I don’t know a damn thing about how house wiring functions except what I have been told by people who do know how house wiring functions, in so much as anyone can know how house wiring functions given that a lot of houses are patchwork monsters of wiring systems from various eras. John had stopped pushing the question on the adapters only because he thought he had figured out how they worked.

All of which led me to give the Doctor a mental high-five for the line from the new season, “I try never to understand. It’s called an open mind.”

And that’s the kicker at the heart of science, isn’t it? Not knowing. Anyone who stayed awake long enough to pass their 101 science requirements in undergrad knows that you can’t ever prove a theory, not really. You can repeat results that fit with the theory enough times that people get comfortable with the notion that you’re right, but at the end of the day, science is about hacking away at the wrong possibilities, which are plentiful. And whether we’re talking about Pluto, or p-values, or socks, or ground adapters, I think what science does best is remind us that none of us knows nearly as much as we’d like to think we know.

Out on the Open Mind

Still with me? Bless your patient heart, and here’s the tl;dr decoder ring at the bottom of the serial (see what I did there, eh, eh?) box…

  1. Unknowing stuff is hard.
  2. Other people can help us unknow stuff,
  3. But sometimes they don’t know that they don’t know stuff either.
  4. This is fine, because nobody really knows stuff.
  5. So what can we do?
  6. Mostly, just keep in mind that we don’t know stuff, and, of course…

Keep looking.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.