When the holiday season comes around, a slew of incredibly important decisions come with it... like choosing which type of cookies to buy. You might find yourself standing in the grocery store, debating between two options: one cookie is less expensive, healthier, and you know it’s a fan favorite. But the other one is your favorite. It just wouldn’t feel like the holiday season without this admittedly very average cookie, though you know there could be better options out there. So, you decide to go with the classic choice and think to yourself that maybe next year you’ll finally switch it up.

Even with an example as simple as holiday shopping choices, it’s clear that the study of human decision-making can be quite complicated. Decision theory is a cross-disciplinary field that examines how individuals, groups, and systems make choices under conditions of uncertainty or complexity. The theory draws on research from psychology, economics, statistics, and neuroscience to understand and model how decisions are made, and is divided into two primary branches: normative and descriptive decision theory.

Normative decision theory focuses on how decisions should be made according to rational criteria, typically under assumptions of perfect information and utility maximization. The objective, then, is to model the ideal decision process. For example, expected utility theory is a foundational model in normative decision-making that suggests that people should make choices that maximize their expected utility by weighing the likelihood of various outcomes against their respective benefits.^1,2

In contrast, descriptive decision theory explores how decisions are actually made, proving that human behavior often diverges from normative standards due to cognitive limitations, biases, emotions, and heuristics (mental shortcuts). For example, prospect theory, developed by Daniel Kahneman and Amos Tversky, describes how people are more sensitive to losses than gains, revealing that our decisions are influenced by more than just pure probabilities, but also by our subjective feelings about their potential outcomes.¹

Utility: The level of satisfaction, value, or benefit an individual derives from a particular outcome. It quantifies the desirability of different choices, allowing for comparison to identify the option with the highest perceived value. Utility serves as the basis for making rational decisions aimed at maximizing personal benefit.

Expected Utility Theory: A model in normative decision-making, suggesting that people should make choices that maximize their expected utility by weighing the likelihood of various outcomes against their respective benefits.³

Preference: An individual's tendency to favor one option over others based on perceived utility. It reflects the subjective ranking of choices, where individuals select options that align best with their goals, needs, or values. Understanding preferences helps explain choice behavior in decision-making models.

Risk: Decision-making in conditions where the probabilities of different outcomes are known. In such situations, individuals weigh the potential gains or losses associated with each choice, often using probability and expected utility to guide their decisions. 

Uncertainty: Situations where outcomes are not fully predictable, and probabilities are unknown or undefined. In decision theory, it complicates decision-making because individuals cannot rely on precise calculations to assess outcomes, often resorting to heuristics or subjective judgments instead.

Prospect Theory: Daniel Kahneman and Amos Tversky’s theory that reveals people are more sensitive to losses than gains, a phenomenon known as "loss aversion." It demonstrates that people don’t always behave rationally and that their decisions are often influenced by how choices are framed or presented.

Bounded Rationality: A concept proposed by Herbert Simon that suggests that decision-makers work within constraints such as limited information, time, and cognitive capacity. This results in "satisficing" rather than optimizing choices.

The history of decision theory stems from philosophical inquiry, mathematical modeling, and psychological experimentation over the last few centuries. 

Early Foundations in Philosophy and Probability (1600s-1700s)

Since decision theory fundamentally relies on probabilities, its roots can be traced to the emergence of probability as a mathematical concept in the 17th and 18th centuries. Blaise Pascal and Pierre de Fermat were among the first to systematically study probability, though originally to solve pressing matters regarding gambling.¹ Their work laid the foundation for thinking about decisions in terms of their likely outcomes. 

Pascal is also known for the eponymous Pascal’s Wager, a philosophical argument for belief in God based on potential outcomes. His wager illustrated an early concept of decision-making under uncertainty, weighing the benefits and risks associated with different choices. The idea of "expected utility" emerged in the 18th century, explaining that people make decisions to maximize not just monetary value but their expected satisfaction, or utility, from an outcome. This led to the expected utility theory framework, which remains central to normative decision theory.^1,4

Rise of Normative Models and Rational Choice Theory (1800s-1940s)

By the 19th century, economics embraced principles of rational choice, which assumed that individuals make decisions to maximize their own benefit or utility. The concept of "homo economicus" (Latin for economic man) portrayed humans as rational beings making logical decisions based on available information.

John von Neumann and Oskar Morgenstern formalized expected utility theory in their 1944 book, Theory of Games and Economic Behavior, which pioneered game theory and set the groundwork for understanding strategic interactions in competitive and cooperative scenarios. The mathematical disciplines of probability and statistics blossomed during this time, providing formal tools for calculating risk and making optimal decisions under uncertainty.^1,4

The Behavioral Revolution and Bounded Rationality (1950s-1970s)

In the mid-20th century, decision theory underwent a major shift with the introduction of behavioral decision theory, which challenged the idea that humans are fully “rational” decision-makers. Psychologists and behavioral scientists began exploring how cognitive limitations and emotions influence decisions. Psychologist and economist Herbert Simon introduced the concept of bounded rationality in the 1950s, arguing that people make "satisficing" decisions (“good enough” rather than “optimal”) because they’re limited by information, time, and cognitive capacity.^1,4

In 1979, psychologists Daniel Kahneman and Amos Tversky developed prospect theory, a landmark decision theory that demonstrated that people make decisions based on perceived gains or losses rather than final outcomes, particularly because they are more hurt by losses than they are uplifted by equivalent gains. This insight into "loss aversion" and the influence of framing on decisions fundamentally changed how economists and psychologists understood human behavior.

Modern Decision Theory and Behavioral Economics (1980s-Present)

In the late 20th and early 21st centuries, the field of behavioral economics emerged, applying insights from psychology to economics. Behavioral economists, including Richard Thaler, built on Kahneman and Tversky's work, emphasizing that real-world decisions often deviate from rational choice models.

Nudge theory, introduced by Thaler and Cass Sunstein, further extended decision theory by exploring how subtle changes in the way options are presented can “nudge” people toward certain choices. This nudging concept has been applied in fields ranging from public policy to personal finance. Advances in neuroscience have also contributed to decision theory, as researchers continue to uncover more about the brain functions underlying decision-making, exploring areas like impulse control, reward processing, and emotional regulation.^1,4

Current Trends and Emerging Approaches

Today, decision theory continues to evolve, integrating findings from data science, neuroscience, machine learning, and artificial intelligence. Predictive models, often powered by large datasets, aim to anticipate human decision patterns—exactly what decision theory strives to do. As artificial intelligence improves and is integrated into more aspects of our lives, we will be able to incorporate larger data sets and arrive at more accurate decision predictions. Of course, understanding and predicting human behavior comes with some challenges, including ethical considerations. Researchers are currently examining issues like algorithmic bias in AI and the ethics of influencing choices through digital platforms.⁵

Blaise Pascal

A 17th-century French mathematician and philosopher, Pascal significantly influenced decision theory with his idea of ‘Pascal's Wager.’ He argued that, under uncertainty about God’s existence, a rational person should "bet" on belief, as the potential benefits outweigh the risks—an early approach to decision-making under uncertainty. His correspondence with Pierre de Fermat about how game winnings should be doled out introduced key principles of probability, crucial for assessing risks and making informed decisions under uncertainty.

John von Neumann

A Hungarian-American mathematician, von Neumann co-authored Theory of Games and Economic Behavior, establishing the foundations of game theory. His work introduced expected utility theory, which became a core model for rational decision-making under risk and uncertainty.

Oskar Morgenstern

An Austrian economist, Morgenstern collaborated with von Neumann on Theory of Games and Economic Behavior, creating a formal framework for decision-making in competitive and strategic situations. His contributions helped shape the study of economics and rational choice theory.

Herbert Simon 

American psychologist and economist, Simon challenged traditional decision theory by introducing the concept of bounded rationality, which suggests that human decision-making is constrained by limited information, cognitive limitations, and time constraints. This concept shifted decision theory from idealized rationality to a more realistic view, acknowledging human limits in processing information.

Daniel Kahneman

A psychologist and Nobel laureate, Kahneman co-developed prospect theory with Amos Tversky, exploring how people make decisions under risk and how biases like loss aversion influence choices. His work shifted decision theory towards understanding systematic deviations from rationality.

Amos Tversky

A cognitive psychologist, Tversky partnered with Kahneman to develop prospect theory and study cognitive biases. His research revealed predictable ways in which human decisions deviate from rational models, transforming theories of decision-making and influencing behavioral economics.

Richard Thaler

An economist and key figure in behavioral economics, Thaler studied how psychological biases and irrational behaviors impact economic decisions. His work on mental accounting, loss aversion, and the endowment effect reveals how real human behavior often deviates from rational models. Thaler’s research, as described in his book Nudge with Cass Sunstein, emphasizes designing choice environments that improve decision-making without limiting freedom.

Cass Sunstein

A legal scholar, Sunstein is known for his work on nudging alongside Thaler. His work highlights the role of choice architecture, which shapes how decisions are presented and perceived, influencing outcomes. Sunstein’s work shows how decision theory can be applied to policy and law to encourage beneficial behaviors in public health, finance, and environmental protection.

Decision theory and the study of human decision have had a widespread impact across numerous fields, from healthcare and marketing to criminal justice and finance, shaping both our understanding of individual choices and institutional interventions. 

Artificial Intelligence and Algorithm Design

Computer algorithms and AI seem to work in mysterious ways. Still, at the heart of their functioning is decision theory, which helps design algorithms that can "think" under uncertainty. This is a crucial feature for many of the major technologies we use today like autonomous vehicles, recommendation systems, and medical diagnostics. By building models that weigh probabilities and outcomes, AI can make choices that mimic rational decision-making, improving safety and personalization in technology. Successfully integrating human decision theory insights into AI could also shape algorithms that align more closely with human intuition and ethics.⁵

Economics and Policy-Making

As we’ve discussed, decision theory provides frameworks like expected utility theory and prospect theory that aid in predicting how individuals and markets will respond to different economic policies. Governments and organizations use these models to design policies that maximize welfare or incentivize desired behaviors, such as encouraging savings or reducing risky investments. By understanding common decision biases, policymakers can implement nudges that guide people toward better financial choices without restricting their freedom.²

Behavioral Finance and Investment

On their own, humans are not always very good at assessing financial risk and determining the best investment strategies. The study of decision theory has shown how biases like overconfidence, loss aversion, and herd behavior can impact investor choices, leading to hugely impactful market anomalies like bubbles and crashes. Recognizing these patterns, many investment firms now incorporate behavioral insights to create tools and strategies that help clients make more rational, stable financial decisions when managing portfolios and trading stocks.⁶

We’ve discussed the differences between normative and descriptive decision theory, which examine how decisions should be made and how they are actually made, respectively. So, which should we use? Should decision theory seek to prescribe ideal choices or accurately represent real-world behavior? The divide between these approaches to decision theory raises philosophical, practical, and methodological questions, as each approach has distinct aims and implications.

Normative Decision Theory 

Because normative theories focus on defining “ideal” decision-making processes, they often use mathematical models to prescribe optimal choices. These theories assume that a rational decision-maker will act in a way that maximizes utility, adheres to probability rules, and minimizes risk based on known information. In fields like economics and policy-making, normative theories are often preferred for creating models and systems that assume “rational” actors, which can simplify predictions and decision-making frameworks.²

However, these often fail to capture the complexity of real-world behavior. We can’t always know what the exact outcome of our decisions will be, and we often don’t have all the facts. The idealized, mathematical models may not always succeed in real-world contexts, where people often face incomplete information or lack the cognitive resources to make purely rational choices. 

Another reason for concern arises when we ask: who defines rationality? Critics argue that, just as different cultures have different values, there is plenty of room for debate on what constitutes appropriate or “rational” choices. If each person’s utility curve varies—and we know that not everyone prioritizes things in the same way—then the optimal choice won’t look the same for everyone. Limiting our concept of rationality may be inherently biased toward those with the power to decide what the right or wrong choices are in a given situation. 

Descriptive Decision Theory

Descriptive theories, on the other hand, aim to observe and model how people actually make decisions, acknowledging the role of cognitive biases, heuristics, and psychological factors. Behavioral researchers like Kahneman and Tversky have shown that individuals frequently deviate from normative models due to biases like loss aversion, framing effects, or social norms. Descriptive models thus recognize that human decision-making is influenced by biases, emotions, social context, and bounded rationality. In areas like psychology and behavioral economics, this perspective is crucial for understanding actual human behavior, which can improve interventions and policies aimed at realistic outcomes rather than idealized ones.¹

However, this empirical focus on actual behavior can lead to overly complex or context-dependent models, potentially making it difficult to derive universal principles or useful predictive insights. This can complicate crafting policy or make potential interventions seem weaker, as the impacts on behavior may appear less certain. 

This perspective may also be limited in applicability, as descriptive theories often study decisions in controlled, low-stakes environments (like laboratory experiments) that might not scale to high-stakes or complex decisions involving multiple factors. The limited scope of many descriptive studies can make it difficult to apply these theories effectively in real-world scenarios that involve even higher levels of uncertainty and risk.

Just as normative decision theory poses the potential for individual or cultural bias, descriptive models can hinder establishing universal patterns or norms. What might be an effective heuristic in one cultural or situational context may not apply in another, leading to difficulties in developing universally applicable decision-making models, and posing a challenge to generating policy or large-scale interventions. 

Hybrid Approach  

Because of the challenges of each approach, some researchers advocate for integrative models that combine elements of both perspectives, using normative theories as a baseline while accounting for deviations observed in descriptive studies. This "bounded rationality" perspective allows for models that assume rationality within limits. Others propose dual-process theories that distinguish between "fast" (intuitive) and "slow" (deliberative) decision-making, blending normative rules with insights from descriptive research.

Decision Theory Programming & Self-Driving Cars

If you’re frequently in the driver’s seat, you might not even be aware of the amount of choices you make in order to arrive safely at your destination. Decision theory allows autonomous vehicles to make complex judgments that can resemble human reasoning but, unlike human drivers, they rely on rigorously programmed decision-making rules governed by risk assessment, probabilistic modeling, game theory, ethical frameworks, and route optimization. Although we will never be able to program the ‘correct choice’ for every possible situation that could occur on the road, the ultimate goal is for self-driving cars to respond safely and intelligently to the vast range of conditions they encounter, and decision theory can help steer us in the right direction.

The first way that self-driving cars use decision theory is in their utility assessment. Since the cars are programmed to maximize safety, efficiency, and passenger comfort, certain utility functions must be programmed, assigning varying utility values to factors like minimizing travel time, avoiding collisions, and maintaining smooth driving. For example, if a car has to decide whether to overtake a slower vehicle, it assesses the potential gains (e.g., saving time and getting you to work quicker) against the risks (e.g., the possibility of a collision... probably not getting you to the office at all). Utility assessment is also used in its route-planning algorithms; when determining the best route, the car evaluates a range of factors, including distance, traffic density, road conditions, and safety. The car’s algorithms will compute the expected utility of different actions and choose the one that maximizes overall utility based on preprogrammed priorities, which will almost always be safety first.^7,8

Roads are inherently unpredictable environments. Self-driving cars use probabilistic decision models to make decisions under uncertainty, constantly updating their estimates based on real-time data from sensors all over the car. For instance, if a pedestrian is approaching a crosswalk, the car doesn’t know exactly when or if they will step onto the road. The decision-making algorithm assesses the probability of that event and weighs potential actions (such as slowing down or stopping) against their probable outcomes. In this case, a pedestrian waiting at the intersection is pretty likely to cross the road, and the car should come to a stop.⁷

Interactions with other drivers or pedestrians can also involve game theory, a branch of decision theory. For example, when merging into traffic, the car must consider how nearby drivers are likely to react. Game-theoretic models enable it to predict the intentions and likely actions of human drivers based on prior interactions and typical driving behaviors, choosing a strategy that minimizes the risk of collision while achieving its goal (e.g., merging smoothly, a challenging task for anyone). This type of decision-making is crucial for "social driving," as self-driving cars have to work cooperatively with other road users, interpreting subtle cues like slight steering or speed adjustments.⁸

In more complex scenarios, decision theory helps self-driving cars navigate tricky ethical decisions. These are the types of edge cases you may have heard about in popular thought experiments, where the vehicle faces a choice between two suboptimal outcomes. A car could be faced with an unavoidable collision scenario, having to choose between staying on course and crashing into a school bus full of children or veering away and crashing into one cyclist. How the car responds would be based on the car’s decision-making system, which may prioritize minimizing harm based on utilitarian principles or pre-set ethical guidelines. Advanced decision models are often coupled with ethical frameworks, allowing the car to analyze potential outcomes in milliseconds, and make choices that reflect programmed values and ethical considerations.⁷ However, the ethics remain tricky to quantify when it comes to potential life-or-death scenarios.

Decision Theory in Medicine

Decision theory plays a critical role in healthcare by enhancing diagnostic and treatment planning processes, allowing doctors to make more informed, evidence-based decisions even in uncertain conditions. Bayesian decision models, for example, are particularly valuable, as they allow clinicians to update the probability of a diagnosis as new information becomes available, such as emerging symptoms or the latest lab test results. For example, if a patient’s symptoms could align with multiple possible conditions, Bayesian models enable the physician to weigh the likelihood of each condition and determine which diagnosis is most probable given all the available evidence.⁹

Moreover, decision theory emphasizes the importance of cognitive bias awareness in clinical settings. Biases like availability bias (where recent or memorable cases disproportionately influence decision-making) and confirmation bias (where clinicians may favor information that affirms an initial diagnosis) can lead to diagnostic errors or suboptimal treatments. By acknowledging these biases, doctors can use structured decision aids, like diagnostic checklists and clinical decision support systems, which help to guide their reasoning and ensure a more balanced evaluation of the patient's condition.

Incorporating decision theory principles into healthcare thus enables clinicians not only to make probability-based decisions but also to account for the psychological factors that might otherwise lead to errors. Ultimately, this approach promotes better health outcomes by improving diagnostic accuracy, optimizing treatment plans, and reducing the risk of misdiagnosis or overtreatment.

The Power of Narratives in Decision Making

Decision-making processes are often described as objective, step-by-step procedures. Humans like to categorize things that way because we tend to process the world around us through sequential narratives with a clear beginning (a cause), middle, and end (an effect). This is known as the theory of narrative thought, and in this article, contributor Constantin Huet explores why it is that we think in terms of stories, and what effect stories have on our consumer decisions. 

Algorithms for Simpler Decision-Making 

To make algorithms more compatible with human judgment, this article advocates for "fast-and-frugal" heuristic-based algorithmic designs that integrate with human intuition, fostering collaborative, participatory human-algorithm decision-making systems.

Rational Actor Theory 

Rational actor theory assumes that people act according to self-interest, choosing the option that maximizes their benefits and minimizes their costs. This article explains the history, consequences, and some of the potential shortcomings of the rational actor model.

1. Steele, K., & Stefánsson, H. O. (2020). Decision theory. In E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy (Winter 2020 Edition). Retrieved from https://plato.stanford.edu/archives/win2020/entries/decision-theory/
2. Briggs, R. A. (2023). Normative theories of rational choice: Expected utility. In E. N. Zalta & U. Nodelman (Eds.), The Stanford Encyclopedia of Philosophy (Winter 2023 Edition). Retrieved from https://plato.stanford.edu/archives/win2023/entries/rationality-normative-utility/ 
3. Biglieri, E. (2022). Beyond probability. In Dimensions of Uncertainty in Communication Engineering (pp. 203-251). https://doi.org/10.1016/B978-0-32-399275-6.00015-0
4. Mao, Felix. (2023). A brief history of decision theory. SGFER. Retrieved from https://www.sgfer.org/wp-content/uploads/2024/01/A_Brief_History_of_Decision_Theory-2.pdf
5. Horvitz, E. J., Breese, J. S., & Henrion, M. (1988). Decision theory in expert systems and artificial intelligence. International Journal of Approximate Reasoning, 2(3), 247–302. https://doi.org/10.1016/0888-613X(88)90120-X 
6. Akin, I., & Akin, M. (2024). Behavioral finance impacts on US stock market volatility: an analysis of market anomalies. Behavioural Public Policy, 1–25. doi:10.1017/bpp.2024.13 
7. Yuan, K., Huang, Y., Yang, S., Zhou, Z., Wang, Y., Cao, D., & Chen, H. (2024). Evolutionary decision-making and planning for autonomous driving based on safe and rational exploration and exploitation. Engineering, 33, 108–120. https://doi.org/10.1016/j.eng.2023.03.018 
8. Waymo. (2024). Waymo driver. Waymo. Retrieved November 1, 2024, from https://waymo.com/waymo-driver/
9. Kornak, J., & Lu, Y. (2011). Bayesian decision analysis for choosing between diagnostic/prognostic prediction procedures. Statistics and its interface, 4(1), 27–36. https://doi.org/10.4310/sii.2011.v4.n1.a4