Decisions Compare Values

Choosing between unlike options creates a peculiar problem.

One job offers a higher salary. Another offers more autonomy, a shorter commute and work that feels more meaningful. None of these properties can be compared directly. Money is not measured in the same units as time, freedom or meaning. Yet a decision still has to make them comparable enough for one option to gain priority.

This does not require a person to translate every feature into a precise internal number. It does require the decision process to represent what the options currently mean for selection.

That represented significance is decision value.

Value is not an intrinsic property of an option

An option does not carry one fixed decision value into every situation.

The same job may appear attractive when financial security is urgent and less attractive when autonomy becomes more important. A convenient meal may have little appeal in one state and dominate the alternatives when someone is exhausted. An immediate reward may gain influence when delay feels uncertain, even if the delayed option would otherwise be preferred.

Decision value is therefore not identical to:

• moral worth;
• market price;
• conscious liking;
• objective benefit;
• permanent preference.

It is the represented significance of an option for selection in the current decision episode.

That representation can incorporate expected reward, effort, delay, uncertainty, social consequences, identity relevance and other attributes. It may reflect long-established preferences and learned associations, but its current influence also depends on the person's state, attention and decision context.

Value-based-choice frameworks describe self-control in similar terms. Instead of treating control as a separate force that suppresses desire, they model choices as competition among outcomes carrying different kinds of value. Immediate reward is one influence, but anticipated consequences, social meaning and identity-related goals can also affect what gains priority. These frameworks do not establish that every decision uses one universal calculation. They establish a more limited point: outcomes must acquire enough represented significance to compete for selection (Berkman et al., 2017; Berkman, Livingston and Kahn, 2017).

Comparison does not require conscious arithmetic

People can often choose without being able to state exactly how they combined the relevant attributes.

Someone may say that one job "just feels better overall." This does not show that no comparison occurred. It shows that the comparison cannot necessarily be recovered as an explicit sequence of conscious calculations.

Some influences may be verbally accessible. A person may deliberately compare salaries or commuting time. Others may operate through learned affective responses, anticipated effort, familiarity or associations that are difficult to articulate. The final judgment can reflect both.

The language of value is therefore functional. It describes the relative influence an option acquires within a choice process. It does not imply a visible mental scoreboard or a person consciously assigning numbers to every possible consequence.

Decisions can unfold through accumulation

Many decisions take time because the evidence favouring each option is incomplete, variable or in conflict.

Evidence-accumulation models describe choice as a process in which information supporting different responses is sampled over time. In these models, selection occurs when the accumulated evidence reaches a decision boundary. Strong and consistent evidence tends to produce faster, more stable choices. Closely balanced or noisy evidence often produces slower and more variable responses.

These models can account jointly for what people choose and how long they take to choose. But their variables must be interpreted carefully. A boundary, drift rate or accumulated value is a model component—not proof that the mind consciously performs the corresponding calculation.

Decision time is not a transparent measure of difficulty either. A slow choice may reflect close competition, cautious responding, distraction, uncertainty or a deliberate demand for more evidence. A fast choice may reflect a large value difference, familiarity, habit or a low decision boundary. Similar response times can therefore arise from different processes.

The models provide structured explanations of choice dynamics. They do not reveal a single mechanism behind every decision.

Attention changes what enters the comparison

A decision cannot be influenced equally by information that is not equally processed.

When one option or attribute receives more attention, information associated with it can make a greater contribution to the evolving comparison. Eye-tracking studies have found systematic relationships between fixation and choice, and attentional drift-diffusion models can reproduce patterns in which gaze and preference develop together (Krajbich, Armel and Rangel, 2010; Krajbich et al., 2012).

This relationship is not one-way. People may look longer at an option because it is already attractive, and looking can also give its attributes more opportunity to influence the decision. Attention and value can reinforce one another during deliberation.

A final choice may consequently depend not only on which attributes were available, but on which were noticed, revisited and processed long enough to affect the comparison. This does not make the choice unreal. It shows that represented value depends partly on how information enters the decision process.

Attributes can be compared in different ways

Not every model assumes that all attributes are first combined into one complete value for each option.

One family of models proposes that salary, autonomy, commute and meaning contribute to an integrated option value. Evidence favouring each option then accumulates until one reaches a decision boundary.

Another family models the accumulation of differences between attributes. The process may repeatedly compare how the options differ on salary, then autonomy, then commuting time, with attention and weighting determining how strongly each difference affects the developing choice.

Other accounts propose smaller ordinal comparisons. Instead of calculating how much better an attribute is on an absolute internal scale, the system may repeatedly register whether a value is better or worse than sampled alternatives.

These models can sometimes predict similar choices while offering different explanations of how comparison occurs. They should not be treated as interchangeable, and current evidence does not establish one representational format for every type of decision.

What they share is narrower: unlike features must acquire some form of relative influence before a choice between them can be made.

The choice set can change the comparison

If every option possessed one stable value independent of its surroundings, adding an irrelevant alternative should not alter the ranking of the original options.

Under some experimental conditions, however, it does.

Return to the two jobs. One offers high pay but limited autonomy. The other offers lower pay but substantially greater autonomy. Now introduce a third position that pays slightly less than the high-salary job while offering even less autonomy. Although this third option is unlikely to be chosen, its presence can make the original high-salary job appear more attractive by providing a nearby alternative that it clearly dominates.

This is an example of a contextual preference effect. It does not occur in every decision, and different models explain such effects differently. Its importance lies in what it rules out: observed preference cannot always be understood as a direct reading of values that remain unchanged across choice sets.

Decision by Sampling offers one account of this context sensitivity. It proposes that an attribute's subjective impact can depend on ordinal comparisons with values sampled from the immediate environment and memory. A salary, delay or effort level may therefore be represented differently when the comparison sample changes (Stewart, Chater and Brown, 2006).

Context does not make value arbitrary. Stable constraints, learned priorities and objective differences still matter. But their influence is expressed through a comparison process that can depend on which alternatives are present.

Preferences may become more definite while deciding

A person does not always enter a decision with a complete ranking waiting to be reported.

Some attributes may initially matter only vaguely. Their importance can become clearer as attention moves between options, consequences are imagined and conflicts become explicit. Research on value certainty suggests that deliberation can refine attribute representations unevenly: attributes receiving more weight may become more precisely represented than attributes that remain peripheral (Lee and Hare, 2023).

This means that a decision can partly reveal an existing preference and partly make that preference more definite.

Claims of preference construction still require methodological care. When people rate options before choosing and then rate the selected option more highly afterward, the difference can appear to show that the choice changed their preferences. Some of that pattern can instead result from noise in the first ratings: the later choice may reveal information that the initial measurement failed to capture. Choice can genuinely contribute to later evaluation, but rating changes alone do not prove that preference was created by the act of choosing (Chen and Risen, 2010).

The careful conclusion is not that preferences are either fixed or invented. Their expression can combine existing dispositions with information refined during the decision itself.

A common currency is a limited hypothesis

How can the system compare outcomes as different as food, money, social approval and relief from effort?

One influential proposal is that different rewards are converted into a common neural currency. Studies of intertemporal choice have found activity correlated with modelled subjective value, and a coordinate-based meta-analysis of 206 functional-imaging studies identified recurring positive value-related signals in areas including the ventromedial prefrontal cortex and anterior ventral striatum (Kable and Glimcher, 2007; Bartra, McGuire and Kable, 2013).

These findings support partly overlapping value-related representations across reward categories. They do not establish that every kind of value is encoded identically, that one region contains the value of an option, or that the brain possesses a single decision centre.

Functional magnetic resonance imaging measures changes in blood oxygenation associated with neural activity. Correlations between those signals and modelled subjective value do not identify a complete neural code, and recurring activation in the same broad region does not prove that every task uses the same underlying representation.

"Common currency" is therefore useful when treated as a constrained hypothesis about partial comparability. It becomes misleading when it is taken to mean one conscious scale, one universal format or one location in which decisions are made.

A choice reveals less than it appears to reveal

When someone chooses the higher-paying job, the observable act establishes that the option gained priority in that episode. It does not establish why.

Salary may have dominated the comparison. Financial uncertainty may have increased its influence. The person may have attended more closely to immediate consequences than distant ones. The alternatives may have changed how each attribute was represented. Several influences may have converged on the same response.

The same person may choose differently later without either choice being false. Their circumstances, knowledge, attention or state may have changed. The choice set itself may be different.

Observed behaviour remains essential evidence, but it does not disclose a permanent internal ranking. It reveals the outcome of a particular comparison under particular conditions.

Selection is not completed action

A decision process can end with one response selected while the corresponding action never occurs.

The person may choose the more meaningful job but fail to submit the application. They may decide to decline an immediate reward but encounter an unexpected opportunity before acting. A selected intention may lose priority when the environment or internal state changes.

This distinction matters because theories of decision and theories of action initiation address related but non-identical problems. Explaining why one option won the comparison does not by itself explain whether the person could initiate, sustain or complete the selected behaviour.

Choice identifies current priority. Action requires that priority to survive contact with the conditions of execution.

Decisions compare current worlds

A decision transforms unlike possibilities into competitors.

It may integrate attributes, accumulate differences, sample comparisons or use partly overlapping value-related representations. Attention determines which information contributes. Context changes the field in which attributes are judged. Deliberation can make initially uncertain preferences more definite. State and learning alter the influence that possible outcomes acquire.

None of this means that choice is arbitrary. It means that value is represented in relation to a person, a situation and a set of alternatives.

Decisions do not simply uncover one permanent value hidden inside each option. They construct a comparison among possible worlds and allow one option to gain current priority.