Resting Bitch: How your moods make you a more optimal you

“Happiness depends not on how well things are going…but whether they are going better than expected.”

Opinion papers are fun. It’s like going to a bar with a scientist and asking what they really think. They might be wrong, but there are very few people far out enough on that ledge to even be spitballing here, so they slap the “opinion” label on and start riffing away. Here, in this churn of conjecture, where bits of evidence are sized up like puzzle pieces, is where new hypotheses are born. Here is the engine that makes science go.

I’ve been dying to write about a recent opinion paper in Trends in Cognitive Sciences. The paper, called Mood as Representation of Momentum, is by Eran Eldar, Robb Rutledge, Raymond Dolan, and Yael Niv, a group from University College London and Princeton. In it, they attempt to bring together two bodies of research that don’t usually interact much. The first area of research focuses on the causes of moods and what happens when they go awry in disorders like anxiety, depression, and bipolar disorder. The second focuses on reinforcement learning (learning from rewards, which can roughly be thought of as trial-and-error learning) and decision making.

Moods, for our purposes, are similar to emotions, but longer-lasting and less specific. Your mood can be an “up” state or “down” state, depending on whether you are happy (in a good mood) or sad (in a bad mood). Your moods then make you more or less likely to experience more specific emotions: for instance, a bad mood can make it easier for you to become angry or frustrated or both.

Moods can be affected by all sorts of things, music, social interaction, self-reflection, sunshine, or even simply viewing the facial expressions of others. In labs, psychologists and economists use pictures of emotional faces, monetary rewards, or even pictures of the outcomes of sporting events to try to manipulate mood. Using smartphone apps, alarms for set reflection times, daily mood journals, brain scans, and more, scientists are attempting to understand what causes moods and what they are for.

***

The authors write: “The upshot of this research is that mood induced by a stimulus can affect judgments about other, potentially unrelated, stimuli. Indeed, this property may have given mood its reputation as a rich fountain for irrational behavior.”

Irrational, indeed. I may have blood coming out of my wherever when I go to bat for my right to mood, but I am not alone. The authors in believe that moods are actually evolutionarily advantageous, that far from being irrational or counterproductive, moods serve a purpose. This argument is a tough sell: moods are typically associated with mood swings, explosive tempers, and generally being a bitch. Coincidentally, many rationality fetishists have a pesky misogyny problem.

To me, this gender-related distaste for moodiness reeks of generation after generation of men taught to bury their feelings. Feelings bad. Boys don’t cry. To aspire to be master of the universe is to aspire to an unattainable objectivity, to become some sort of stoic thetan, freed from the sway of irrational emotional forces.

Toxic masculinity remains a top rant for me. So I was pretty dang thrilled to see badass Princeton computational modeler Yael Niv arguing that we have evolved to be moody because moods make us optimal learners. That’s right: mood is a feat of evolutionary engineering, a Goddess-given engine of practical efficiency for all people of all genders. Finally, evo psych in the service of something I can get behind. Niv, whose work I know best among the authors, is an expert in feelings, if ever you could call someone one. Her work creates and tweaks actual formulae for happiness. Equations. Plug and chug and Happiness = X. How’s that for rational?

***

Let me back up for a moment. I’ll get to the formula for happiness momentarily. But first, I want to talk about a shirt I once saw. It was emblazoned with a molecular structure and the slogan “Dopamine: technically the only thing you like.” This, while clever, is not technically true. You also like opioids (like morphine and your body’s natural equivalent, endorphins) and arguably serotonin (the chemical behind the chemical imbalance that is depression, and a common target of antidepressant medications), along with who knows what other mysterious neurotransmitters exist but that we have yet to understand. Dopamine, however, is more like the only thing (that we know of) that you want. It drives craving.

When you receive an unexpected reward, you get a burst of dopamine in an area of the brain known as the ventral striatum. A pet peeve of mine is when people show brain responses to cocaine and then do a side-by-side of whatever it is they’re arguing acts similarly. Sugar. Video games. Gambling. I myself spent most of grad school programming an Atari-like shuffleboard game which, though primitive, robustly “lit up” the ventral striatum just like cocaine. This doesn’t mean I’ve created a cohort of shuffleboard addicts. All it means is that this mechanism, this burst of dopamine that signals to you that your expectations have been exceeded, is a very general mechanism.

Now, mind you, this reward must be unexpected in order for it to change your future decision-making behavior. If all you get is the reward you expected, your expectations go unchallenged and you aren’t learning anything, per se. In fact, studies have shown that when monkeys expect a reward and then do not receive one, their dopamine neurons skip a beat, ceasing their firing as though in indignation. Activation in the dopamine-rich ventral striatum has also been measured in humans using functional magnetic resonance imaging (fMRI) in response to all kinds of pleasurable stimuli. And if you give people a pharmacological boost in dopamine, they report greater happiness from rewards than they normally would. Dopamine, in addition to feeling good, makes you want more dopamine. And, helpfully, it’s critical for teaching you how to get it.

In labs, this is sometimes studied by people playing a game where they can either win or lose money. Scientists are interested in the role rewards can play in sculpting three main things: people’s subjective reports of happiness, the brain response to future rewards, and the effect of these rewards in sculpting subsequent decisions. Based on these three measurements, computational modelers can design algorithms that can accurately predict people’s feelings of happiness, brain responses, and decision-making behavior.

For example: people report greater feelings of happiness after winning. These rewards also lead to future rewards having bigger impact on their subsequent decisions–they may, for instance, feel themselves on a hot streak and take bigger risks. Similarly, losing money reduces feelings of happiness. It also reduces the impact future rewards have on their choices, and furthermore, it measurably dampens the brain response to these rewards. Negative events throw a bucket of cold water on us, making us pessimistic.

These patterns are exacerbated for people who are less emotionally stable, suggesting that the study of how people learn from rewards may offer clues to the origins of mood disorders. They may also help explain why these disorders cause people to make the decisions they do. For instance, when people are asked to choose between a sure bet and a risky gamble, with varied gains and losses, their decisions help train an algorithm to produce a model of happiness. These algorithms calculate happiness as a function of their choices (sure bets or risky gambles, or in other words, choosing or avoiding risk), the expected payoff of the gamble, and the difference between the actual and expected payoff. Throw in some weighting variables and constants, like a “forgetting factor” that determines the relative influence of more recent events and events further in the past, and you’ve got an actual formula for happiness.

Scientists have used these types of formulae to make inferences from data acquired in smartphone-based field studies. These found that, despite what you’ve heard about the power of positive thinking, it’s not so much your expectations that impact happiness and learning. These are determined far more directly by the surprise you experience about the outcomes of your decisions. These surprises are also known as prediction errors: the difference, or error, between your predicted outcome and the actual outcome. Happiness can be calculated as a running average of recent reward prediction errors, where some prediction errors are weighted more heavily than others. And wouldn’t you know it: by modeling happiness quantitatively, you can search for activity fitting this model in fMRI scans. And this approach to looking for the seat of happiness will pay off: you will find it right there in that needy ratcheter-upper of need, the ventral striatum.

***

Mood, as we said, can be biased by all sorts of things. Just seeing frowny faces can bias your perception of subsequent rewards. More seriously, being depressed can mean that future rewards have less of an impact on your choices. You are de-sensitized to the meaning of these rewards. You don’t see it, because the part of you that values these rewards has been blunted. Critically, though, it’s not necessarily because your learning is impaired. This is a motivational issue, not accessible to the realm of rational appeal.

Anxiety, like depression, enhances responses to aversive stimuli: you respond to events as if they are worse than they really are. While depression manifests as a greater sensitivity to negative outcomes, positive mood can enhance risk-taking in lab settings as well as in financial markets. A positive mood biases the perceived likelihood of future positive outcomes–in other words, you see everything as coming up roses. Repeated positive prediction errors, or good surprises, can “invigorate reward-seeking behavior.” Dopamine craves dopamine. Good surprises have a way of making you believe that there are lots more good things to come.

That’s because reinforcement learning is all about tracking which states were rewarding and making choices that will get you back there. Think of choosing a “good” slot machines, or, in the wild, of animals seeking out the trees that have the most fruit. Scientists believe they are on to something in using slot machines to test people’s decision making, because when people play this type of game, their behavior seems to pick out optimal strategies.

Mood, the authors argue, smooths out inefficiencies in reward learning. Going back to the example of the animals searching for food in the trees, they write: “Increased rainfall or sunshine may cause fruit to become more abundant in all trees simultaneously. In this situation, it makes little sense to update expectations for each tree independently.” Mood, in the landscape of learning how to reliably reap rewards, is the rising tide that lifts all ships. You don’t want to be constantly surprised to find fruit–this would not be advantageous. Instead, you want to infer that something bigger is going on. So your mood helps you to ratchet up your expectations more quickly than you normally would by allowing all the happy surprises to have an even bigger impact on how much you learn from subsequent happy surprises.

Mood gives you a way of calibrating your learning apparati to account for multiple factors in your environment, without having to account precisely for each one and its impact individually. This form of generalization, far from being sloppy, actually improves the efficiency of learning when multiple reward sources are interdependent, which is pretty much the norm. It’s rare that good things happen in your life that have nothing to do with anything else you did, except maybe winning the lottery. Unless you are that powerball winner, it would be irrational to ignore the connections and treat these things as independent. 

The authors write: “Indeed, such interdependencies may be the rule rather than the exception, for both animals and humans, because success in acquiring skills, material resources, social status, and even mating partners can be tightly correlated.” I’ve watched enough of the Real Housewives of Orange County to tell you that this is not necessarily true, nor would having it all necessarily lead to an improved mood.  

But still, I wonder: Is the otherwise-iffy idea that women are better planners and multitaskers rooted in a real perception of this mood-driven increase in efficiency? Is this form of mathematically generalizing over all currently relevant sources of reward what gives us our supposedly innate abilities to plan? Do men do themselves a disservice by faking such an even keel, and do women suffer disproportionately from anxiety and depression due to waves of dopaminergic dysfunction?

Just wild speculation going on here, by the way. Really going off book. I’m just saying, ignoring or suppressing moods and emotions is probably not the best practice. That’s the only real flame war I am down to start here. Moods are useful! What the authors argue is that if you infer a positive momentum from an increase in the availability of fruit in your orchard, you may be on to something: spring is coming. Same goes for the negative momentum as we head into winter. Better hibernate. By adjusting your expectations as quickly as you need to to catch up the rate of rewards, you save yourself from perpetual shock, perpetual disappointment.

As a quadriplegic man I read about said, “you can get used to anything.”

***

Did everyone see Inside Out? If not, lookout, spoiler. Do you remember when Sadness touched all the memories? And we learned that Riley’s Disgust, Fear, Anger, and Sadness were just as important in guiding her through life as her Joy? Well, so, too, are all sources of information useful in learning how to navigate our environments. Given a good enough probabilistic model of environment, plus some Bayesian magic, you can come up with an optimal learning algorithm for a particular environment.

You want this algorithm to be able to account for environmental factors that are general enough to affect multiple states, or situations, instead of treating all the states as independent. Sure, you might over-generalize sometimes: maybe the increase in fruit is local to the trees in this particular valley. But you weight how recent and how local the changes are to try to account for this as much as you can. You assume that neighboring states have been changed in similar ways to the ones you’re currently being surprised by. Even if you’re not integrating over multiple states, or multiple sources of reward, this generalization can occur over time, too, allowing you to infer momentum from your running average of how many good surprises you seem to be getting lately.

If emotional reactions have an appropriate intensity and duration, then mood is helping you out. Good and bad moods should only stick around as long as there’s still a change in momentum registering–that is, as long as you’re being surprised and having to adjust your expectations. But once your expectations are updated and seem to be in line with your new normal, your happiness levels should reach a more neutral place. Similarly, if you keep encountering bad outcomes, you will get in a bad mood but your expectations will level off appropriately eventually (You can get used to anything). The authors point out that happiness levels return to baseline even after winning the lottery, which is maybe why they say money can’t buy you happiness.

***

But what happens if you have a mood disorder? These can be serious. Excessive happiness or sadness would lead to behaviors that are maladaptive. If you learn less from negative surprises than you do from positive ones, you develop an overly optimistic expectation, which means you’re slammed harder by the negative surprises. High expectations lead to low mood.

A mood that keeps pace appropriately with changes in the environment acts as a homeostatic mechanism to keep your learning processes on track. It’s when your mood is out of sync with the rate of change in the environment that you might run into trouble.

People with depression are thought to have some dysfunction in regulating the levels of the neurotransmitter serotonin in their brains. Low serotonergic function has been known to lead people to learn less quickly from negative outcomes. Depression may result in (or from) negative outcomes having less of an impact on behavior. Normal feedback loops get out of whack, with expectations falling further and further behind realities. To avoid perpetual disappointment, expectations need to be adjusted to match outcomes. But as mismatches grow, bigger mood swings can result. These oscillations may form the basis for bipolar disorder, causing expectations and mood to pitch wildly up and down even when nothing in the environment is changing.

Interestingly, in the general population, positive mood & risk aversion predominate. Risk aversion can make you happy to have what you have, in a good mood as long as nothing is going wrong. This predominance may arise because people learn more, in general, from negative surprises than positive ones, changing their decisions and expectations more markedly when things get bad than when they are going well. This keeps people happy in the face of unpleasantness. The stronger biasing effect that negative outcomes have is likely due to the greater evolutionarily adaptive significance of learning quickly from negative momentum. In other words, it’s more important for our survival to avoid negative outcomes than to maximize the positive ones. If you don’t find the most fruit, you’ll probably be fine, but if you don’t learn to run from predators, you’re dead.

***

I come from a long line of pessimists and worriers. I can’t help but think that sensitizing oneself to negative outcomes is a helpful form of vigilance that is maybe not such a bad character trait. But so where do we land on the power of positive thinking versus setting one’s expectations low so that you will never be disappointed? Well, it’s telling that if you treat people with major depressive disorders by giving them serotonergic drugs, their perceptions change before their mood does. In other words, putting on the rose colored glasses comes first, and seems to be the cause of the improvement in mood. So what antidepressants are really giving you is not a direct mood boost, but rather a shift in your perception that results in one. How can you do this without drugs? Who knows. If I knew that, I wouldn’t be writing this stuff for free.

Bottom line: Mood can sensitize (or de-sensitize) you to the outcomes of your decisions, increasing (or decreasing) your responsivity to them. Emotional instability could, in theory, arise from either moods having too strong a sensitizing effect or from weakening people’s ability to habituate to new normals. The evidence suggests that people who are emotionally unstable tend, if anything, to show stronger effects of outcomes on their feelings (they are sensitive) and a stronger influence on their evaluation of subsequent outcomes. Their hair-trigger reflex for inferring momentum may lead to overgeneralization, and inappropriate optimism or pessimism.

But without this generalization, come on. We’d be simple idiots. We’d be rats bar-pressing for our rewards. We’d be stuck in a railbound behaviorist hellscape of rote stimulus-response associations, Skinner boxes made of skin and bone.

***

The authors clinch an important win for moody bitches everywhere by closing the paper with: “Moods can reflect inference of momentum even when there is none in the environment, leading to excessive optimism or pessimism. However, the ubiquity of moods and the extent of their impact on our lives tells us that, throughout the course of evolution, our moodiness must have conferred a significant competitive advantage. Being moody at times may be a small price to pay for the ability to adapt quickly when facing momentous environmental changes.”

Give me, then, the power of mentally smearing the causal influence of many unrelated outcomes together, or give me death.

Advertisements
This entry was posted in Angry Woman Stuff, Science, Uncategorized. Bookmark the permalink.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s