Expected Value and Expected Utility in Decision Science

Last Updated June 4, 2026

Expected value and expected utility are foundational concepts in decision science that provide formal methods for evaluating choices under uncertainty. Expected value calculates a probability-weighted average of outcomes, while expected utility incorporates subjective preference, allowing decision-makers to account for risk attitudes and differences in perceived value.

This article is part of the Decision Science knowledge series.

These concepts form part of the analytical core of decision theory and remain central to decision science more broadly. They provide a structured way to compare alternatives when outcomes are uncertain, enabling decision-makers to move beyond intuition toward explicit and systematic evaluation.

At the same time, their interpretation and application require care. Expected value can be mathematically clear yet behaviorally misleading. Expected utility can model risk preference more realistically, yet still depend on assumptions about probability, preference stability, and the meaningfulness of a utility scale. In practice, these tools are most powerful when treated not as automatic answer machines, but as disciplined ways of clarifying what is uncertain, what is valued, and how trade-offs are being made.

Painterly editorial illustration of expected value and expected utility with dice, weighted outcomes, branching probability paths, tradeoff scales, curved value forms, and reflective decision-makers. — Expected value compares probable outcomes, while expected utility accounts for how people or institutions value those outcomes under uncertainty.

Expected value: a probabilistic foundation

Expected value (EV) is the simplest formal method for evaluating uncertain outcomes. It represents the weighted average of all possible outcomes, where each outcome is multiplied by its probability of occurrence. In mathematical terms, it converts a probability distribution over outcomes into a single summary quantity.

The appeal of expected value lies in its clarity. If probabilities are known and outcomes can be expressed in a common unit, the option with the highest expected value can be identified as the best choice for a risk-neutral decision-maker. This makes expected value especially useful in gambling, finance, engineering reliability, insurance, and other domains where repeated exposure or portfolio logic can make average outcomes meaningful.

However, expected value treats outcomes only in terms of their numerical magnitude. It does not ask whether a gain of 100 means the same thing to a poor decision-maker as to a wealthy one, or whether a large downside risk is psychologically or practically tolerable even when the average payoff is attractive. That limitation is one reason expected utility became so important.

Expected utility: incorporating preferences and risk

Expected utility (EU) extends expected value by introducing the idea that outcomes are not evaluated purely as objective quantities. Instead, they are transformed through a utility function that reflects the decision-maker’s subjective valuation of those outcomes.

This move is historically associated with Daniel Bernoulli’s attempt to resolve the St. Petersburg paradox. Bernoulli argued that people do not respond to money linearly and that the value of an increment of wealth depends on the level of wealth already possessed. That insight made it possible to explain why real decision-makers would reject gambles with extremely high expected monetary value if the structure of the gamble imposed unacceptable risk or diminishing marginal benefit.

Expected utility therefore allows decision-makers to model different attitudes toward risk:

Risk-averse: preferring more certain outcomes even when expected value is lower
Risk-neutral: focusing only on probability-weighted average outcome
Risk-seeking: preferring higher-variance outcomes with potentially larger gains

This makes expected utility more behaviorally and normatively flexible than expected value alone. It also explains why expected utility became a core part of modern decision analysis.

Utility functions and decision behavior

The shape of the utility function determines how uncertainty is valued. A concave utility function implies diminishing marginal utility and therefore risk aversion. A linear function implies risk neutrality. A convex function implies risk-seeking behavior over the relevant range.

In principle, utility functions allow highly nuanced decision modeling. In practice, however, they are not always easy to elicit or justify. Preferences may vary across individuals, domains, time horizons, and institutional settings. A person may be risk-averse about health outcomes, risk-neutral in routine budgeting, and risk-seeking in entrepreneurial strategy.

These complications show that utility is powerful as a formal tool but not automatically simple as an empirical one. Real-world choice often involves multiple objectives, moral constraints, and qualitative considerations that cannot always be reduced cleanly to a single scalar function without loss.

The St. Petersburg paradox and Bernoulli’s intervention

The St. Petersburg paradox occupies a central place in the history of expected utility because it exposed the inadequacy of expected monetary value as a universal rule of rational choice. The gamble has an unbounded expected monetary value, yet ordinary decision-makers are not willing to pay an arbitrarily large amount to enter it.

Bernoulli’s 1738 essay, later translated as “Exposition of a New Theory on the Measurement of Risk,” proposed that the value of money should be understood through utility rather than nominal magnitude. This was one of the decisive conceptual shifts in the history of decision theory because it replaced a purely monetary criterion with a subjective valuation criterion.

That shift still matters. It shows that rational evaluation under uncertainty cannot always be reduced to arithmetic expectation. It must also account for how outcomes are experienced, not only how they are counted.

Limits of expected value and expected utility

Expected value and expected utility are powerful, but both have important limitations. First, they require probabilities. In many real-world situations, probabilities are uncertain, disputed, unstable, or impossible to estimate with confidence. When the probability model itself is fragile, both EV and EU become harder to apply cleanly.

Second, expected utility assumes that preferences can be represented coherently and with enough stability to support formal analysis. Yet behavioral research shows that preferences are often context-dependent, influenced by framing, loss aversion, reference points, and cognitive bias.

Third, these models typically evaluate outcomes in a relatively contained way. In complex systems, outcomes may depend on feedback loops, interdependence, dynamic adaptation, and path dependence. In those environments, the decision problem may not be adequately represented by a fixed menu of states and payoffs alone.

From expected utility to modern decision science

Modern decision science builds on expected value and expected utility rather than abandoning them. These concepts still provide the baseline grammar of formal choice under uncertainty. But contemporary practice often extends them through sensitivity analysis, scenario comparison, Bayesian updating, robust decision-making, and multi-criteria evaluation.

Instead of relying solely on one expected value or utility calculation, decision-makers increasingly examine how recommendations change across different assumptions, models, and future conditions. This shift is especially important in environments characterized by deep uncertainty, contested probabilities, and system complexity.

In that sense, expected utility remains foundational but no longer stands alone. It is one element in a wider decision architecture that must also address model uncertainty, learning, behavior, and resilience.

Applications in decision-making

Expected value and expected utility remain central in many applied domains:

Finance: portfolio analysis, pricing, and risk management
Healthcare: treatment choice, diagnostic decision analysis, and screening policy
Engineering: reliability analysis, safety trade-offs, and contingency design
Public policy: cost-benefit reasoning, resource allocation, and policy evaluation

In each of these domains, these concepts help structure uncertainty into comparable terms. But their usefulness depends on whether they are embedded in broader frameworks that account for uncertainty quality, model assumptions, stakeholder values, and practical constraints.

Behavioral critiques and descriptive departures

One of the major developments in modern decision science is the recognition that expected utility does not always describe how people actually choose. The work of Tversky and Kahneman, along with later behavioral economics, showed systematic deviations from expected-utility predictions, especially under framing, loss aversion, and probability distortion.

This does not make expected utility obsolete. It clarifies its role. Expected utility remains a powerful normative and analytical tool, but descriptive theories are often needed to explain real human behavior under uncertainty. The contrast between expected utility and behavioral decision theory has become one of the defining tensions in the field.

That tension is productive. It allows decision science to distinguish between how choice can be analyzed formally and how it is actually enacted psychologically.

Implications for decision science

Expected value and expected utility have several enduring implications for decision science:

Uncertainty can be formalized: outcomes under risk can be compared systematically rather than impressionistically
Preferences matter: value is not exhausted by raw monetary or numerical magnitude
Risk attitude matters: the same gamble may be rationally treated differently under different utility structures
Formal clarity has limits: probabilities, utility, and system structure all need scrutiny

These implications explain why EV and EU remain indispensable. They are not the whole of decision science, but they are among its deepest foundations.

Mathematical Lens: Expected value, expected utility, and risk attitude

The expected value of an uncertain prospect with outcomes \(x_1, x_2, \dots, x_n\) and probabilities \(p_1, p_2, \dots, p_n\) is:

\[
EV = \sum_{i=1}^{n} p_i x_i
\]

This expression is appropriate when outcomes are valued linearly and the decision-maker is effectively risk-neutral.

Expected utility replaces raw outcomes with a utility transformation \(u(x)\):

\[
EU = \sum_{i=1}^{n} p_i u(x_i)
\]

If \(u(x)\) is concave, the decision-maker is risk-averse over the relevant range. A common illustrative utility function is logarithmic utility:

\[
u(x) = \ln(x)
\]

or, more generally, a constant-relative-risk-aversion form:

\[
u(x) = \frac{x^{1-\rho} – 1}{1-\rho}, \qquad \rho \neq 1
\]

where \(\rho\) governs the degree of risk aversion. The larger \(\rho\), the more sharply large risky gains are discounted relative to certainty.

This formal distinction captures the conceptual difference between EV and EU: expected value asks what happens on average, while expected utility asks how the decision-maker values the average structure of uncertainty.

Advanced R Workflow: Comparing Risk-Neutral and Risk-Averse Choices

The R workflow below compares stylized uncertain prospects under expected value and log-utility scoring. It illustrates how a choice ranking can change when subjective risk attitude is introduced.

# Install packages if needed:
# install.packages(c("tidyverse"))

library(tidyverse)

# ------------------------------------------------------------
# R Workflow: Comparing Risk-Neutral and Risk-Averse Choices
# Purpose:
#   Compare stylized prospects under expected value
#   and expected utility using log utility.
# ------------------------------------------------------------

prospects <- tibble(
  prospect = c("Safe Option", "Balanced Gamble", "High-Risk Gamble"),
  outcome_1 = c(100, 180, 400),
  prob_1 = c(1.0, 0.6, 0.25),
  outcome_2 = c(100, 40, 0),
  prob_2 = c(0.0, 0.4, 0.75)
)

prospects <- prospects %>%
  rowwise() %>%
  mutate(
    expected_value = outcome_1 * prob_1 + outcome_2 * prob_2,
    expected_utility = prob_1 * log(outcome_1 + 1) + prob_2 * log(outcome_2 + 1)
  ) %>%
  ungroup()

print(prospects)

prospects_long <- prospects %>%
  select(prospect, expected_value, expected_utility) %>%
  pivot_longer(
    cols = c(expected_value, expected_utility),
    names_to = "metric",
    values_to = "value"
  )

ggplot(prospects_long, aes(x = prospect, y = value, fill = metric)) +
  geom_col(position = "dodge") +
  labs(
    title = "Expected Value vs Expected Utility",
    x = "Prospect",
    y = "Value",
    fill = "Metric"
  ) +
  theme_minimal(base_size = 12)

write_csv(prospects, "expected_value_expected_utility_profiles.csv")

Advanced Python Workflow: Simulating Utility Curves and Choice Under Uncertainty

The Python workflow below simulates how different levels of risk aversion alter the utility assigned to uncertain prospects. It illustrates how the same probabilistic structure can yield different rankings depending on the utility function.

# Install packages if needed:
# pip install pandas numpy matplotlib

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# ------------------------------------------------------------
# Python Workflow: Simulating Utility Curves and Choice
# Under Uncertainty
# Purpose:
#   Compare stylized prospects across different
#   risk-aversion parameters.
# ------------------------------------------------------------

prospects = {
    "Safe Option": [(100, 1.0)],
    "Balanced Gamble": [(180, 0.6), (40, 0.4)],
    "High-Risk Gamble": [(400, 0.25), (0, 0.75)]
}

risk_levels = [0.2, 0.8, 1.4]
results = []

def crra_utility(x, rho):
    x = x + 1e-6
    if abs(rho - 1.0) < 1e-9:
        return np.log(x)
    return (x ** (1 - rho) - 1) / (1 - rho)

for rho in risk_levels:
    for name, outcomes in prospects.items():
        expected_value = sum(payoff * prob for payoff, prob in outcomes)
        expected_utility = sum(prob * crra_utility(payoff + 1, rho) for payoff, prob in outcomes)
        results.append({
            "risk_aversion": rho,
            "prospect": name,
            "expected_value": expected_value,
            "expected_utility": expected_utility
        })

df = pd.DataFrame(results)
print(df.head())

plt.figure(figsize=(10, 6))
for rho in risk_levels:
    subset = df[df["risk_aversion"] == rho]
    plt.plot(subset["prospect"], subset["expected_utility"], marker="o", label=f"rho={rho}")

plt.xlabel("Prospect")
plt.ylabel("Expected Utility")
plt.title("Choice Under Uncertainty Across Risk Aversion Levels")
plt.legend()
plt.tight_layout()
plt.show()

summary = df.groupby("prospect").agg(
    average_expected_value=("expected_value", "mean"),
    average_expected_utility=("expected_utility", "mean")
).reset_index()

print(summary)
summary.to_csv("utility_curve_choice_summary.csv", index=False)

Conclusion

Expected value and expected utility provide the formal foundations for evaluating decisions under uncertainty, transforming uncertain outcomes into structured comparisons. Expected value offers a clean probabilistic benchmark, while expected utility introduces subjective valuation and risk preference, bringing analysis closer to the realities of choice.

These concepts remain essential to decision science, but they must be applied in ways that acknowledge their assumptions and limits. Their deepest value lies not in providing automatic answers, but in clarifying how uncertainty, consequence, and preference are being combined when a decision is made.

References

Bernoulli, D. (1954) ‘Exposition of a new theory on the measurement of risk’, Econometrica, 22(1), pp. 23–36. Translation of the 1738 essay. Stable bibliographic discussion available at: Stanford Encyclopedia of Philosophy.
Howard, R.A. and Abbas, A.E. (2023) Foundations of Decision Analysis. Harlow: Pearson. Available at: Pearson.
Raiffa, H. (1968) Decision Analysis: Introductory Lectures on Choices Under Uncertainty. Reading, MA: Addison-Wesley. Bibliographic record available at: Google Books.
Tversky, A. and Kahneman, D. (1974) ‘Judgment under uncertainty: Heuristics and biases’, Science, 185(4157), pp. 1124–1131. Available at: Science.