IN THE MOOD|
Neuroscientist Takes a Gamble on Happiness
Are you happy now? That probably depends on many things including recent events, your choices and expectations. As circumstances change over time, so might your mood. Given certain variables, can we predict how emotions might fluctuate?
Neuroscientist Dr. Robb Rutledge is studying what determines our emotional state, specifically whether we can predict mood changes over time. This research may provide insights into the emotional states and decisions of people with depression, bipolar disorder and other mental health conditions.
Rutledge, principal research associate at University College London’s Max Planck Centre for Computational Psychiatry and Ageing Research, built a mathematical model to sum up happiness in terms of rewards and expectations (see below). Think of it as a roll of the dice. The building excitement of whether you’ll win a bet and the resultant gain or loss can mimic the oscillating moods inherent in certain mental disorders.
“I think these computational approaches are very powerful for understanding subjective states that are relevant to psychiatry,” said Rutledge at a recent NIMH Director’s Innovation Speaker Series lecture at the Neuroscience Center. “These tools may help us bridge the gap between neurobiology...and subjective symptoms assessed by clinicians.”
Rutledge wanted to test his hypotheses on a large data set. For his purposes, clinical trials had limited reach; it’s expensive to collect data from thousands of patients. So Rutledge helped develop a free smartphone app called the Great Brain Experiment (www.thegreatbrainexperiment.com), which features cognitive and emotional tasks in the form of games, to augment the research of smaller clinical trials and laboratory studies.
“One thing we can do with smartphones is cheaply collect data among large populations that can help us to answer questions that are important for psychology and psychiatry,” Rutledge said. To date, more than 120,000 people have used the app.
In one game, for example, the player can choose the safe option or spin a wheel to win points, then rates his or her happiness based on the outcomes. The player makes 30 choices and 12 happiness ratings in a 4-minute game.
“We’re finding that happiness depends not on how well people are doing in the task, but whether they’re doing better than expected,” Rutledge said.
His team has replicated this finding in tens of thousands of subjects based on data from the app, which also confirmed results from their smaller lab studies that used brain imaging to show that neural activity in dopamine-related brain areas predicted changes in happiness. This surprised Rutledge because, he said, people playing the app games are anonymous, unpaid and might not answer honestly about their happiness.
“We can think about these tasks that we’re running as probes of the neural circuits that underlie psychiatric symptoms,” Rutledge said. “One advantage of putting patients in tasks [on their phones] is we can measure how they behave and feel in environments that a clinician wouldn’t be able to assess, such as a situation in which they face multiple negative events in a row. We can control their environment because we set up [the game] that way.”
A key component in Rutledge’s happiness equation is the reward prediction error, the difference between what people expect and what they receive. In our brains, dopamine neurons signal reward prediction errors. Getting an unexpected positive reward boosts dopamine activity, whereas getting a predicted reward doesn’t make us quite as happy and keeps dopamine at baseline.
Data from 1,800 players of the Great Brain Experiment, which included more than 400 people diagnosed with depression, showed that reward prediction errors had similar effects on fluctuations in mood in depressed and non-depressed subjects.
“These results suggest that the dopamine system that produces reward prediction errors is probably functioning normally in depression and that the reward-related symptoms of depression have a different cause,” said Rutledge.
You win some, you lose some, but what happens to your mood when your friend wins or loses? In another experiment, Rutledge said, participants were less happy if they won and another player lost. The player’s happiness really took a dive if he or she lost and the other player won. “Both forms of inequality—getting more or less than another person—are happiness-reducing,” said Rutledge.
Mood changes might also lead to a potentially dangerous feedback cycle. Low expectations make positive surprises larger, which lead to inflated expectations that set the stage for negative surprises and lower mood—a cycle of alternating periods of positive and negative emotion common to those with bipolar disorder. Rutledge’s lab is building a computational model to understand the interaction between emotional states and resultant decisions, including risky decisions typical in bipolar patients.
As Rutledge was collecting data from his app, he realized that many players only used it on the day they downloaded it, which didn’t provide the longitudinal data he sought for clinical questions. This spring, his lab will launch a follow-up app with new cognitive science games that will provide data over time by asking people to play the games every couple of weeks and fill out brief questionnaires about their current mood.
The four games focus on risk, effort, learning and other behaviors relevant to depression. In one game, you’re a fisherman who earns different amounts of points per fish and the values change over time. In another game, you’re digging for treasure and know that you will have winning and losing streaks down the line.
“We’re interested in how information about future prospects affects your current emotional state and the decisions you make,” said Rutledge.
The smartphone games provide hundreds of data points reflecting the player’s emotions and behaviors. The computational models then reduce these data to a much smaller number of parameters that can be used as inputs to machine-learning algorithms that can make predictions about symptoms.
“To the extent they actually capture important aspects of those neural circuits,” said Rutledge, “we expect they would be useful for making clinical predictions, for predicting whether symptoms will get better or worse.”