The ice cream tastes fine. You know it does. Objectively, this is good ice cream. But somewhere between your tongue and whatever part of your brain is supposed to light up like a pinball machine, the signal just... fizzles. Welcome to anhedonia, the clinical term for when your reward system goes on indefinite lunch break.

A new comprehensive review in Neuroscience and Biobehavioral Reviews tackles one of psychiatry's slipperiest problems: how do you measure something as subjective as "not enjoying stuff anymore"? And can we use math to figure out what's actually broken?

More Than Just Sad - The Wanting Problem

Here's where it gets interesting. Anhedonia isn't just about pleasure being muted - though that's part of it. Modern research has split the concept into multiple components: there's consummatory anhedonia (the ice cream doesn't taste good) and anticipatory anhedonia (you can't get excited about getting ice cream tomorrow). Then there's the motivation piece - even if you'd enjoy the ice cream, you can't summon the energy to walk to the freezer.

According to Cleveland Clinic, roughly 70% of people with major depressive disorder experience anhedonia. It's not a side effect of depression - it's one of the core symptoms, serious enough that you can be diagnosed with depression based on anhedonia alone, even without feeling traditionally "sad."

Teaching Algorithms to Understand Empty

This is where computational psychiatry enters the chat. Researchers have been using reinforcement learning (RL) models - the same mathematical frameworks that help robots learn to play video games - to understand how depressed brains process rewards differently.

The basic idea: our brains constantly make predictions. "If I do X, I'll feel Y amount of good." When reality differs from prediction, that's a prediction error, and healthy brains use these signals to update their expectations. The neurotransmitter dopamine acts like the brain's Yelp review system, flagging when experiences exceeded or fell short of expectations.

In depression, this system malfunctions. Research shows that simulating synapse loss in artificial neural networks produces remarkably depression-like behaviors - including anhedonia, increased impulsivity, and avoidance. The artificial agents started bypassing optional rewards entirely, mirroring what researchers observe in depressed humans who can't be bothered to pursue pleasurable activities.

The Effort Equation

One particularly clever research tool is the Effort Expenditure for Rewards Task (charmingly abbreviated "EEfRT" - pronounced "effort"). It measures whether someone will work harder for bigger rewards. Turns out, people with anhedonia consistently choose easier tasks even when the bigger reward is clearly worth the extra effort. It's not that they can't do the math - they just can't make themselves care enough about the outcome to bother.

Studies on depressed adolescents using reinforcement learning models found that more anhedonic youth showed slower "evidence accumulation" - basically, their brains took longer to process reward-related information and make decisions. The machinery works, just in low-power mode.

What the Models Actually Found

The review examined 37 studies using computational models of anhedonia and found something sobering: most models were decent at describing what's happening (face validity) but pretty bad at predicting who would respond to treatment (predictive validity). They could capture the cognitive and behavioral stuff but struggled with the underlying neurobiology.

This matters because if we want to develop better treatments - and current options for anhedonia are frustratingly limited - we need models that can tell us which brain circuits to target. The good news? Cognitive behavioral therapy appears to actually improve the reinforcement learning parameters in depressed patients. After treatment, their brains showed higher reward learning rates and stronger signals in the ventral striatum. The system isn't permanently broken - it can be recalibrated.

Building Better Mental Models (Of Mental Models)

The authors propose something ambitious: an integrative, systems neuroscience approach that treats anhedonia as a multi-dimensional phenomenon spanning reward processing, executive function, and even sensory processing. Think of it like debugging software - you can't fix a complex bug by only looking at one function. You need to trace the problem through the entire system.

For those of us who enjoy making sense of complex information, tools like mapb2.io can help visualize how these interconnected research findings relate to each other - because honestly, mapping out the relationships between reward circuits, dopamine signaling, and behavioral tasks is exactly the kind of thing your brain wants to see drawn out.

The Bigger Picture

What makes this research genuinely hopeful isn't that we've solved anhedonia - we haven't. It's that we're finally asking the right questions with tools sophisticated enough to find meaningful answers. Instead of just noting that depressed people "feel less pleasure," we can now quantify how their reward learning differs, which computations are impaired, and potentially why certain treatments work.

The brain that forgot how to want things isn't fundamentally different from yours. It's running the same algorithms, just with different parameters. And parameters, unlike personality traits or life circumstances, are things we might actually be able to adjust.

References

• Singh, S., Cunningham, J.E.A., Uher, R., Becker, S., & Nunes, A. (2026). The conceptualization, measurement, and critical appraisal of computational models of anhedonia in depression. Neuroscience and Biobehavioral Reviews. DOI: 10.1016/j.neubiorev.2026.106652

• Treadway, M.T., et al. (2009). Worth the 'EEfRT'? The Effort Expenditure for Rewards Task as an Objective Measure of Motivation and Anhedonia. PLOS ONE. DOI: 10.1371/journal.pone.0006598

• Chen, C., et al. (2024). The computational structure of consummatory anhedonia. Trends in Cognitive Sciences. PMID: 38423829

• Simulation of synapse loss inducing depression-like behaviors (2024). Nature Communications. PMID: 39569353

• Computational reinforcement learning, reward (and punishment), and dopamine in psychiatric disorders. Frontiers in Psychiatry (2022).

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.