Sunday, April 17, 2011

Trust but verify: side-effects of a life in science

I had an interesting experience about a week ago while hanging out with a new friend.

We were buying beer and pooling our pocket money to give as a present for the host of a going-away party, and didn't have anything to wrap it in. We wanted to be classy and get an envelope.

The scenario basically went like this. We walked into a small liquor store and bought beer. Then we went up to the counter to pay.

Friend: Do you have any envelopes? Or plain white paper?

Cashier: No.

Friend: Okay, we'll just take the beer.

Then we went and looked around the store again, to see if there were any office supplies we had somehow missed. We couldn't find anything.

Then my friend did the unthinkable.

Friend: Are you sure you don't have any plain white paper? Just one sheet back there behind the counter?

Cashier: Oh, yes. Here you go.

I was dumbfounded. Essentially the same question, asked twice, yielded opposite answers.

(If this happens in science, you go home and drink whiskey and come back the next day and try it 8 more times and take careful notes and try to figure out why you got one answer 5 times and the opposite answer 3 times, and then you modify your protocol and do it another 8 times and if you get the same answer 7 times, you publish it and say +/- 1. And write a long discussion section explaining what you think the variables are.)

In this case, I would never have thought to ask again. I spent the whole week thinking about it.

It reminded me of this time during my postdoc when I was struggling with setting up a new technique and I wasn't sure I really wanted to do it and it didn't work on the first try and it was a total pain the ass so I really didn't want to do it over again.

My advisor gave me two pieces of seemingly conflicting advice:

1. I'm sure you did it right. Don't second-guess yourself.

2. Just try it one more time. If it still doesn't work, move on. Get everything fresh and just try it again before you give up.

Now, statistical gurus will tell you that these two truths are not mutually exclusive. Yes, you could have done it perfectly the first time and just happened to grab a black sock instead of a red one. Yes, if you stick your hand back in the sock drawer and rummage around again, this time you might get a red sock.

But when you're trying to make the all-important decision WHAT DO I DO NOW, sometimes it's hard to feel like you're making the right choice. Are you wasting time by trying again? Are you being impossibly stubborn when you should really try a different approach? Or are you giving up too easily?

Which also reminds me of a time during grad school when I was struggling with a technique that worked initially and then stopped working for no apparent reason. I could not figure out what had changed. I hadn't changed anything.

One of my committee members said: You're near an edge.

And I said, What?

She explained that sometimes you get lucky and something works when it shouldn't. Then you go chasing after what are actually the wrong conditions for an experiment, when you'd be better off chucking the whole thing and starting fresh from first principles.

In this particular case, she was absolutely right. We redesigned the experiment and I got it working, consistently.

I never did get it to work the original way again.

It's a little heartbreaking and mind-fucking when you think you've seen something real, and potentially really interesting, and you can't reproduce it. If you're lucky, you're in a field where there are plenty of things to choose from, and you can pick any one of these to pursue or at least try a few times.... before you give up. And then you take a deep breath and try a different one.

In real life, you don't get all these do-overs.

Sometimes you're dealing with the most unpredictable thing in the universe: other people.

I've noticed that doing science for so long has made me really reluctant to trust my own decision making. Yes, in science I always went with my gut, and when I let people (my advisors) talk me out of doing that, I always regretted it. But it was also a fairly low-risk endeavor. I learned early on how to test the waters with cheap, quick pilot experiments. And I never got too attached to any one experiment (unlike my advisors and the ubiquitous asshole reviewer).

I think this lack of self-trust is actually one of the things that makes a great scientist great. The unwillingness to trust any result, no matter how appealing, until it has been thoroughly verified.

You can't say, "Well, it must be right, I did it! And I am awesome!"

No. You have to say, "This might be wrong."

You have to say it over and over and over. Even if you did it yourself with your own two hands, and saw the result with your own two eyeballs. Even if you think it's the coolest thing in the world and you desperately want it to be published in Cell so you can get a faculty position and a grant and a fleet of minions to do your experiments for you.

You always have to keep in mind that the journey isn't from Maybe Wrong -> Definitely Right.

It's from Uncertainty ---> Close Enough For Now.

On a long enough timeline, everything turns out to be just an approximation of the truth.

Sometimes you're pretty much right and you can be proud of that. But you can't really be sure until hindsight.

In real life, sometimes you have to go with your gut and hope you aren't completely fucking everything up. And sometimes when you ask for what seems like an impossible do-over, you can get one. Just because you asked nicely.

Labels: , , , ,