Breakup Lines, Prediction Errors, and Learning
Dear Amara,
“I think she’s going to be OK,” my boyfriend messaged after breaking up with me by text. “I let her down easy.” “I think you meant to send this to someone else,” I replied.
Michelle Weathers, Laurel, Md.
So New York Times put up this … article? … collection?... called “The 52 Best Breakup Lines (Said in Real Life).” And this one above made me WHOA out loud (wol? Lol) Link to this article can be found at the end of this letter.
Actually, the first time I read this, I read “I will let her down easy” in my head…oof. Shocked much?
Prediction Errors, Learning Rates — What are they, and Why the Kale should you care?
Something about unexpected breakups unceremoniously done gets people, or at least I hope it’s not just me? There’s this thing that psychologist call prediction errors – you have an expectation of how things are gonna go, and things happen that positively or negatively shocks you. Psychologists also think that this prediction error is the basis of learning, and there is another parameter (think system setting that controls how much or little you do something), call the “learning rate”, that controls how fast you learn from your prediction errors.
Too slow, of course, is not good – as the saying that got perhaps mistakenly attributed to einstein goes, “doing the same thing over and over again and expecting different result is insanity” – we want to learn from our mistakes speedily if possible so we can improve. However, updating your beliefs about the world too swiftly also has its downside – you don’t want to “downgrade” a relationship that you’ve had for a long time with ample evidence that it was nourishing on one thing the other party did that irked you in a given moment. In other words, depending on the nature of the event or relationship, calibrating to the optimal learning rate, if there is such a thing, is tricky.
Is it possible to learn not too fast and not too slow?
Why calibrate at all you ask? I don’t know about you but I, for one, don’t really relish the feeling of being let down – quite literally a description of negative prediction errors. I don’t mind positive surprises as much though – a positive prediction error, however, arguably it could also have a potential downside: it raises the bar. Once you got surprised by your partner on your birthday, you perhaps would expect it next year and feel a small pang of disappointment if they didn’t do something quite as nice. Adjusting our expectation to the actual state of the world lessens the likelihood that your stomach will experience some sort of lurch in such events. Perhaps it also lessens the chance to experience excitement. But then as many therapists say, the feeling of excitement and anxiety can feel very similar in the body. And I’m honestly not quite sure if it isn’t a worthwhile trade off, that never experiencing excitement brings you forever peace from anxiety.
Can you be truly rational? Never say never…
But of course, the first rule of science (according to me, today, ask me again tomorrow), is to avoid using words such as “never” and “forever.” So perhaps all that musing is just for naught, and try as we might, we cannot train our brains as if it’s a smart LLM and tune the parameters to optimize our feeling states – notice how paradoxical this sentence is: our brain is much more complicated than an LLM, and there is no such thing as optimizing feeling states when it comes to today’s LLM, and hopefully not in the future either.
Feelings and rationality have been pitted against each other through the ages, with philosophers and psychologist arguing for the superior importance of one or the other or both in conjunction in guiding our behavior. But alas, sometimes human behavior is so good it moves us to tear, sometime it is so evil you’d think whatever guides it must be a faulty mechanism.
Upon reaching the usual confounding conclusion, I bid you good day and ask you one more question:
Here’s the link to the article on 52 real life break up lines, which one are you most “impressed” (maximum positive or negative prediction error) by?
Boo!👻
Ina
