THE MACHINERY OF EFFORT VALUATION
How every action is priced before the person decides.
What follows is not advice about productivity.
It is the mechanism that runs before productivity is even possible.
Every action a person could take gets a price tag. Not after deliberation. Before it. The price is computed by circuits that finish their work before the conscious mind has formed a sentence about what to do.
The person experiences the output as a feeling.
“That seems hard.”
“I don’t feel like it.”
“Maybe later.”
These are not decisions. They are readouts. The verdict was already delivered. The feeling is the receipt.
The computation is not mysterious. It has inputs, weights, a location, a learning rule, and a set of failure modes that explain most of what people call laziness, procrastination, burnout, and depression.
This is what the computation looks like.
PART ONE: THE PRICING MODEL
Every potential action enters a cost function.
The function integrates at minimum eight variables:
THE EFFORT COST FUNCTION
┌───────────────────────────────────────────────────────┐
│ │
│ 1. reward magnitude how much the outcome is │
│ worth │
│ │
│ 2. reward probability how likely the outcome │
│ actually arrives │
│ │
│ 3. temporal delay how long until the │
│ outcome lands │
│ │
│ 4. physical effort metabolic and muscular │
│ cost of the action │
│ │
│ 5. cognitive effort executive function load │
│ the action demands │
│ │
│ 6. opportunity cost what else the same │
│ resources could do │
│ │
│ 7. prior cost estimate learned from every past │
│ encounter with this type │
│ of action │
│ │
│ 8. current energy state interoceptive signals │
│ about available fuel │
│ │
└───────────────────────────────────────────────────────┘
The brain does not weigh these variables equally. It does not weigh them rationally. And it does not subtract them linearly.
The discounting follows a hyperbolic curve.
This means the cost function is steepest at the near end. Delaying a reward by one day from now drops its value far more than delaying it by one day from a month out. Adding ten minutes of effort to an easy task feels much larger than adding ten minutes to a task that already takes three hours.
The mathematical form:
SUBJECTIVE VALUE = REWARD / (1 + k * EFFORT)
k = individual sensitivity to effort cost
┌────────────────────────────────────────────────────┐
│ │
│ value │
│ │ │
│ │ · │
│ │ · │
│ │ · │
│ │ · │
│ │ · │
│ │ · │
│ │ · · · · · · │
│ │ │
│ └──────────────────────────────────────────► │
│ effort demanded │
│ │
│ the curve is steep at the start │
│ then flattens │
│ small early costs destroy value │
│ large late costs barely register │
│ │
└────────────────────────────────────────────────────┘
This curve is not a metaphor. It fits the behavioral data across thousands of experiments. People discount effort hyperbolically, and they discount delay hyperbolically, and these two discounting processes are separable. A person can have steep effort discounting and shallow delay discounting, or the reverse. The brain runs both computations through overlapping but distinct circuits.
The output of the function is a scalar. A single number. The net expected value of acting.
If the number is high enough, the action initiates.
If it is not, the action does not happen. The person does not experience a decision. They experience a wall.
PART TWO: THE PRICING CIRCUIT
The computation has a physical address.
THE EFFORT PRICING CIRCUIT
┌───────────────────────────────────────────────────────┐
│ DORSAL ANTERIOR CINGULATE (dACC) │
│ │
│ the central hub │
│ integrates reward, effort cost, control demand │
│ outputs the expected value of control signal │
│ fires cost prediction errors when effort ≠ expected │
│ │
└───────────────────────────┬───────────────────────────┘
│
┌───────────────┼───────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ ANTERIOR │ │ VENTRO- │ │ VENTRAL │
│ INSULA │ │ MEDIAL PFC │ │ STRIATUM │
│ │ │ │ │ │
│ reads the │ │ encodes net │ │ reward │
│ body's state │ │ subjective │ │ anticipation │
│ heart rate │ │ value after │ │ willingness │
│ glucose │ │ costs are │ │ to exert │
│ fatigue │ │ subtracted │ │ effort │
│ autonomic │ │ │ │ │
│ arousal │ │ the verdict │ │ the fuel │
│ │ │ │ │ line │
│ the body │ │ │ │ │
│ budget │ │ │ │ │
└───────────────┘ └───────────────┘ └───────────────┘
│
▼
┌─────────────────────┐
│ LATERAL HABENULA │
│ │
│ the quit signal │
│ fires when costs │
│ exceed benefits │
│ inhibits dopamine │
│ promotes giving │
│ up │
└─────────────────────┘
The dACC is where the integration happens. It receives inputs from the insula about the body’s current state, from the striatum about reward expectations, and from prefrontal regions about the cognitive demands of the task. It computes a single signal that Shenhav, Botvinick, and Cohen named the Expected Value of Control.
This signal specifies two things simultaneously. How much the action is worth. And how much cognitive control to allocate to it.
If the EVC is high, control resources are deployed. Attention sharpens. The action begins.
If the EVC is low, control resources are withheld. The action stalls. The person reaches for the phone instead.
The vmPFC then encodes the final verdict. The net subjective value of acting after all costs are subtracted. This is dissociable from the dACC’s signal. The dACC mobilizes effort. The vmPFC estimates value. These are different computations performed by different tissue.
And when costs exceed benefits decisively, the lateral habenula fires. It inhibits dopamine neurons through the rostromedial tegmental nucleus. Dopamine drops. The motivational signal collapses.
The person experiences this as giving up.
They call it quitting.
It is a circuit firing.
PART THREE: WHAT DOPAMINE ACTUALLY DOES
The popular understanding is wrong.
Dopamine is not the pleasure chemical.
Decades of work by Salamone, Correa, and others demonstrate that dopamine’s primary motivational function is regulating willingness to exert effort. Not the experience of reward. The willingness to work for it.
The evidence is clean. Animals with depleted mesolimbic dopamine can still experience pleasure. Their consummatory behavior is intact. They eat the food, they enjoy the sugar. But given a choice between a high-effort option that delivers more reward and a low-effort option that delivers less, they shift toward the low-effort option.
They did not stop wanting.
They stopped being willing to work.
DOPAMINE: THE EFFORT WILLINGNESS SIGNAL
┌───────────────────────────────────────────────────────┐
│ │
│ what people think: │
│ │
│ dopamine = pleasure │
│ more dopamine = more happiness │
│ reward releases dopamine │
│ │
│ what the data shows: │
│ │
│ dopamine = effort allocation │
│ more dopamine = more willingness to work │
│ dopamine fires on PREDICTION of reward │
│ not on receipt │
│ │
│ D1 receptors → sensitivity to BENEFITS │
│ D2 receptors → sensitivity to COSTS │
│ opponent channels on the same currency │
│ │
└───────────────────────────────────────────────────────┘
There is a deeper layer.
Tonic dopamine concentration. The baseline level that is not phasic, not spiking in response to any particular event. Niv, Daw, Joel, and Dayan showed in 2007 that this baseline encodes the average reward rate of the environment.
This is the opportunity cost signal.
If the environment is generally rich, tonic dopamine is high. Any action that takes too long or yields too little is penalized, because every moment spent on it is a moment not spent collecting the average environmental return.
If the environment is generally poor, tonic dopamine is low. The threshold for action drops. Even small rewards become worth pursuing, because there is nothing better available.
This is why a person in a stimulating environment (social media, notifications, variable reward schedules) finds boring tasks intolerable. Tonic dopamine is high. The opportunity cost of effort is high. The quiet task cannot compete.
The person does not lack discipline.
Their pricing circuit is correctly computing that the boring task returns less than the environment’s average reward rate.
The problem is not in the person.
The problem is in the environment they have built around themselves.
PART FOUR: THE OPPORTUNITY COST MODEL
The willpower depletion model is wrong.
Baumeister’s ego depletion model proposed that self-control draws from a limited metabolic resource. Use it up and the tank empties. This model has failed to replicate, and the physiology was never plausible. Total brain glucose consumption barely changes between cognitively demanding and idle states.
Kurzban, Duckworth, Kable, and Myers proposed the replacement in 2013.
The feeling of effort is not depletion. It is opportunity cost.
Executive function mechanisms can only be deployed for a limited number of tasks simultaneously. Using them on Task A means not using them on Tasks B through N. The subjective experience of difficulty is the felt readout of this opportunity cost computation.
EFFORT IS NOT DEPLETION. IT IS OPPORTUNITY COST.
┌───────────────────────────────────────────────────────┐
│ │
│ WRONG MODEL (depletion): │
│ │
│ resource ████████████████████ → task → resource ███ │
│ "willpower tank empties" │
│ │
│ CORRECT MODEL (opportunity cost): │
│ │
│ task A gets attention → tasks B, C, D neglected │
│ felt difficulty = signal that B, C, D exist │
│ and may be more valuable than A │
│ │
│ the feeling is a REALLOCATION SIGNAL │
│ not a fuel gauge │
│ │
└───────────────────────────────────────────────────────┘
When a person says “I’m mentally exhausted,” they are not describing a depleted battery. They are describing a system that has computed that the current task’s expected return no longer justifies monopolizing limited executive resources. The system is lobbying for reallocation.
This is why the same person who “cannot” continue working on a report can immediately switch to an engaging conversation, a video game, or a crisis at work. The executive resources are not gone. They are being withheld from the current task because the pricing says the current task is not worth it.
The fatigue is real as a feeling.
It is not real as a description of resource availability.
It is a verdict from the pricing model.
PART FIVE: THE BODY BUDGET
The pricing model does not run in a vacuum.
It receives continuous input from the body about available resources, current demands, and predicted future needs. This input arrives through the anterior insula, which maintains a real-time map of the body’s internal state.
Heart rate. Blood glucose. Cortisol. Muscle fatigue. Sleep debt. Immune activation. Digestive load.
All of these feed into the cost function.
Barrett’s framework names this allostasis. The brain does not wait for homeostatic error signals. It predicts future metabolic needs and adjusts effort allocation proactively. The brain is budgeting the body’s energy before the body sends a distress signal.
THE BODY BUDGET INPUT
┌──────────────────────┐ ┌──────────────────────┐
│ BODY SIGNALS │ │ ANTERIOR INSULA │
│ │ │ │
│ heart rate ────►│────►│ re-represents │
│ glucose level ────►│ │ body signals as │
│ cortisol ────►│ │ subjective state │
│ muscle fatigue────►│ │ │
│ sleep debt ────►│ │ feeds into dACC │
│ immune load ────►│ │ cost function │
│ │ │ │
└──────────────────────┘ └──────────┬───────────┘
│
▼
┌──────────────────────┐
│ EFFORT COST │
│ ESTIMATE │
│ │
│ shifted UPWARD │
│ when body budget │
│ is stressed │
│ │
│ shifted DOWNWARD │
│ when body budget │
│ is flush │
└──────────────────────┘
When the body budget is stressed, every action’s cost estimate rises. Not because the actions changed. Because the metabolic context changed.
This is why everything feels harder after a bad night of sleep. The tasks are the same. The body budget shifted. The cost function repriced every action in the person’s repertoire.
The person thinks they are weak.
They are reading a shifted cost function.
PART SIX: HOW THE PRICES GET SET
Effort costs are learned. Not fixed. Not innate. Learned through a specific mechanism: cost prediction errors.
The brain maintains running estimates of how costly each type of action is. These estimates update via the difference between expected effort cost and experienced effort cost. Just like reward prediction errors update value estimates, cost prediction errors update effort estimates.
There is a critical asymmetry.
The update happens retrospectively. At effort completion. Not when the effort cue is first presented. The brain does not update its effort model when you see the task. It updates the model after you have done the task.
THE COST LEARNING RULE
before action:
┌───────────────────────────────────────────────────┐
│ │
│ stored estimate: "this action costs 8 units" │
│ │
└───────────────────────────────────────────────────┘
after action:
┌───────────────────────────────────────────────────┐
│ │
│ actual experienced cost: 3 units │
│ │
│ cost prediction error: 8 - 3 = 5 │
│ │
│ model update: next estimate drops toward 3 │
│ │
└───────────────────────────────────────────────────┘
the model only learns when the action is TAKEN
the update only fires at COMPLETION
if the action is avoided, no error, no update
This learning rule is the key to the entire machinery.
If the action is taken and the actual cost is lower than predicted, the estimate drops. Next time the action is priced cheaper. Initiation becomes easier.
If the action is taken and the actual cost is higher than predicted, the estimate rises. Next time the action is priced higher. Initiation becomes harder.
If the action is not taken, nothing happens. The estimate persists unchanged.
This last case is where the system breaks.
PART SEVEN: THE AVOIDANCE TRAP
Avoidance is computationally devastating.
When a person avoids an action, no effort is exerted. No cost is experienced. No prediction error is generated. The prior estimate is never corrected.
The brain’s effort cost estimate for the avoided action persists at whatever level it was when avoidance began. If the system has a pessimistic prior, the estimate may drift upward over time.
But there is a second mechanism that makes it worse.
Avoidance generates a relief signal. The predicted cost was high. The action was not taken. The predicted cost did not materialize. This registers as negative reinforcement. The avoidance behavior itself is strengthened.
THE AVOIDANCE LOOP
┌──────────────────────────────┐
│ INFLATED COST ESTIMATE │
│ "this will be terrible" │
└──────────────┬───────────────┘
│
▼
┌──────────────────────────────┐
│ AVOIDANCE │
│ "not today" │
└──────────────┬───────────────┘
│
┌─────────┴──────────┐
│ │
▼ ▼
┌──────────┐ ┌─────────────────┐
│ RELIEF │ │ NO PREDICTION │
│ SIGNAL │ │ ERROR │
│ │ │ │
│ avoidance │ cost estimate │
│ behavior │ never updated │
│ reinforced │ │
└──────────┘ └────────┬────────┘
│
▼
┌──────────────────────┐
│ ESTIMATE PERSISTS │
│ OR DRIFTS HIGHER │
│ │
│ loop repeats │
└──────────────────────┘
This is the mechanism underneath procrastination, avoidance disorders, and the slow narrowing of a life that stops trying things.
The model predicted the action would cost 8. The person avoided it. The model still says 8. Next time the person encounters the same action, the model says 8 again. The avoidance is reinforced by the relief. The price never corrects.
If the person had taken the action, they might have discovered the actual cost was 3. The prediction error would have fired. The model would have updated. Next time the price would have been closer to 3.
But they did not take the action.
So the model never learned.
This is why exposure works. Not because it eliminates fear. Not because it habituates the person. Because it forces the prediction error that avoidance has been suppressing. The goal of exposure is not comfort. It is maximal negative prediction error. The thing was less costly than predicted. The model updates.
PART EIGHT: SEROTONIN AND PERSISTENCE
Dopamine controls initiation. Serotonin controls persistence.
Optogenetic activation of serotonin neurons in the dorsal raphe nucleus increases willingness to wait for delayed rewards and to sustain active effort at a reward site.
Computational analysis by Meyniel and colleagues shows that serotonin’s effect is specific. It diminishes the weight of effort cost in the cost function. It does not increase reward sensitivity. It does not change the reward magnitude variable. It changes the k parameter. The effort sensitivity coefficient.
THE DOPAMINE / SEROTONIN DIVISION
┌───────────────────────────────────────────────────┐
│ │
│ DOPAMINE │
│ controls: effort initiation │
│ action vigor │
│ opportunity cost baseline │
│ mechanism: D1 for benefits, D2 for costs │
│ failure: action never starts │
│ │
├───────────────────────────────────────────────────┤
│ │
│ SEROTONIN │
│ controls: effort persistence │
│ patience across delay │
│ cost weight reduction │
│ mechanism: dorsal raphe → dACC cost signal │
│ failure: action starts but stops early │
│ │
└───────────────────────────────────────────────────┘
This is why SSRIs can help with avoidance behavior. They do not make the reward bigger. They make the effort feel less costly. The k parameter shifts. The denominator in the cost function shrinks. The net value of acting rises.
Not because the world changed.
Because the pricing changed.
PART NINE: HOW THE SYSTEM BREAKS
The system breaks in at least four ways. Each produces a recognizable pattern.
Elevated cost sensitivity. The most distinctive computational feature of depression across multiple paradigms is not reduced reward sensitivity. Reward processing is relatively spared. It is the cost side that is miscalibrated. The effort sensitivity parameter k is elevated. Every action is overpriced. The person can still enjoy things when they arrive. They cannot initiate the effort to reach them.
This is why depressed people are not anhedonic in the strict sense. Hand them a piece of chocolate and they rate it as pleasant. Ask them to walk across the room for it and they cannot start.
Frozen predictions. Chronic avoidance prevents cost prediction errors from firing. The model never encounters disconfirming evidence. Estimates become entrenched. The longer avoidance persists, the more the person’s repertoire of priced actions narrows to only those with the lowest cost estimates. The life shrinks.
Disconnected update circuit. In learned helplessness, the medial prefrontal cortex loses its regulatory control over the dorsal raphe nucleus. The mPFC normally inhibits passive coping responses. When this control fails, the organism defaults to behavioral inaction regardless of actual contingencies. The model may or may not be accurate. It does not matter. The update circuit is offline.
Inflated environmental reward rate. A person who has spent years in a high-stimulation environment (social media, on-demand entertainment, variable reward schedules) has a high tonic dopamine baseline. The opportunity cost of any low-stimulation task is extreme. The quiet task returns less than the average environmental reward rate. The pricing says: not worth it.
FOUR FAILURE MODES
┌──────────────────┬────────────────────────────────┐
│ failure │ what breaks │
├──────────────────┼────────────────────────────────┤
│ depression │ k parameter elevated │
│ │ everything overpriced │
│ │ │
│ chronic │ avoidance suppresses │
│ avoidance │ cost prediction errors │
│ │ model frozen │
│ │ │
│ learned │ mPFC loses control over │
│ helplessness │ passive coping circuit │
│ │ update mechanism offline │
│ │ │
│ stimulation │ tonic dopamine inflated │
│ saturation │ opportunity cost of effort │
│ │ exceeds quiet task value │
└──────────────────┴────────────────────────────────┘
In each case, the person experiences the same thing.
“I can’t make myself do it.”
The statement is honest. They cannot. But the reason is not character, not laziness, not weakness. The reason is a cost function that is outputting the wrong price.
PART TEN: THE DRIFT THRESHOLD
The moment-to-moment decision to act can be modeled as evidence accumulation.
Noisy evidence for “act” versus “don’t act” accumulates over time. A response initiates when accumulated evidence reaches a threshold. This is the drift-diffusion model.
Three parameters matter:
THE ACTION THRESHOLD
┌───────────────────────────────────────────────────────┐
│ │
│ DRIFT RATE │
│ speed at which evidence for "act" accumulates │
│ reduced by high effort cost → slower accumulation │
│ │
│ BOUNDARY SEPARATION │
│ how much evidence is needed before committing │
│ raised by uncertainty → more evidence required │
│ │
│ STARTING POINT BIAS │
│ prior inclination toward acting or not acting │
│ shifted by past experience with similar actions │
│ │
└───────────────────────────────────────────────────────┘
high effort cost =
slow drift rate +
high boundary +
starting point biased toward inaction
result:
longer delay before initiation
higher probability of choosing inaction
the experience of "I can't get started"
Effort costs reduce the drift rate. Evidence for “act” accumulates more slowly. They may also raise the decision boundary. More evidence is required before committing.
The combined effect is increased reaction time, increased probability of choosing the null option, and the subjective experience of difficulty.
This is not a metaphor for what happens in the brain. It is a mathematical description of what the neural populations in the pre-supplementary motor area, the frontal eye fields, and the prefrontal cortex are computing when the person stares at the task and does not begin.
PART ELEVEN: THE SINGLE FACT
The feeling of effort is not a report on reality.
It is a prediction.
The brain generated a cost estimate. The estimate may be accurate. It may be inflated by avoidance, distorted by depression, shifted by the body’s energy state, or corrupted by an environment that has raised the baseline reward rate above what any quiet task can match.
The person does not know whether the estimate is accurate until the action is taken. If the action is taken and the actual cost is lower than predicted, the model corrects. If the action is avoided, the model persists.
THE SINGLE FACT
┌───────────────────────────────────────────────────┐
│ │
│ the feeling of "this is hard" │
│ is a PREDICTION │
│ not a MEASUREMENT │
│ │
│ the prediction can be wrong │
│ │
│ the only way to correct it │
│ is to take the action │
│ and let the prediction error fire │
│ │
│ avoidance does not just delay the action │
│ it prevents the correction │
│ that would make future action easier │
│ │
└───────────────────────────────────────────────────┘
This is the machinery.
It prices every move before the person moves. It learns from experience but only from experience that is actually had. It can be miscalibrated by avoidance, by depression, by stimulation saturation, by metabolic stress, by the slow drift of a life that stopped testing its own predictions.
The machinery does not care about willpower. It does not respect goals. It does not respond to motivation. It responds to prediction errors.
And prediction errors only happen when the action is taken.
Citations
The pricing model and effort discounting:
- Westbrook, A. & Braver, T.S. (2015). Cognitive effort: A neuroeconomic approach. Cognitive, Affective, & Behavioral Neuroscience, 15, 395-415.
- Klein-Flugge, M.C. et al. (2016). Effort discounting functions in physical and cognitive domains. PLOS ONE, 11(7).
Expected Value of Control:
- Shenhav, A., Botvinick, M.M. & Cohen, J.D. (2013). The expected value of control: An integrative theory of anterior cingulate cortex function. Neuron, 79(2), 217-240.
- Shenhav, A., Musslick, S., Lieder, F. et al. (2017). Toward a rational and mechanistic account of mental effort. Annual Review of Neuroscience, 40, 99-124.
Dopamine and effort willingness:
- Salamone, J.D. & Correa, M. (2012). The mysterious motivational functions of mesolimbic dopamine. Neuron, 76, 470-485.
- Salamone, J.D. et al. (2018). Dopamine, effort-based choice, and behavioral economics. Frontiers in Behavioral Neuroscience, 12, 52.
Tonic dopamine and opportunity cost:
- Niv, Y., Daw, N.D., Joel, D. & Dayan, P. (2007). Tonic dopamine: Opportunity costs and the control of response vigor. Psychopharmacology, 191, 507-520.
The opportunity cost model of effort:
- Kurzban, R., Duckworth, A., Kable, J.W. & Myers, J. (2013). An opportunity cost model of subjective effort and task performance. Behavioral and Brain Sciences, 36(6), 661-679.
Interoception and the body budget:
- Craig, A.D. (2009). How do you feel now? The anterior insula and human awareness. Nature Reviews Neuroscience, 10, 59-70.
- Barrett, L.F. (2017). The theory of constructed emotion: An active inference account of interoception and categorization. Social Cognitive and Affective Neuroscience, 12(1), 1-23.
Effort cost learning and prediction errors:
- Botvinick, M.M., Huffstetler, S. & McGuire, J. (2009). Effort discounting in human nucleus accumbens. Cognitive, Affective, & Behavioral Neuroscience, 9, 16-27.
- Mental effort cost learning is retrospective. (2022). bioRxiv.
Temporal discounting:
- Ainslie, G. (2001). Breakdown of Will. Cambridge University Press.
- Zhang, S. et al. (2022). A neuro-computational account of procrastination behavior. Nature Communications, 13, 5723.
Serotonin and persistence:
- Meyniel, F. et al. (2016). A specific role for serotonin in overcoming effort cost. eLife, 5, e17282.
Computational psychiatry of effort:
- Treadway, M.T. et al. (2012). Effort-based decision-making in major depressive disorder. Journal of Abnormal Psychology, 121(3), 553-558.
- Maier, S.F. & Seligman, M.E.P. (2016). Learned helplessness at fifty. Psychological Review, 123(4), 349-367.
- Huys, Q.J.M. et al. (2015). Computational psychiatry as a bridge from neuroscience to clinical applications. Nature Neuroscience, 18, 1561-1571.
Drift-diffusion models:
- Ratcliff, R. & McKoon, G. (2008). The diffusion decision model: Theory and data for two-choice decision tasks. Neural Computation, 20(4), 873-922.