Psychology of Learning

Module 09: Decision-Making 1

Part 3: Models of Decision-Making

Looking Back

In Parts 1 & 2, we explored choice behavior from empirical & theoretical perspectives. The Matching Law describes how organisms distribute responses to match reinforcement distribution. Behavioral economics explains this pattern through optimization theory. We distinguished decision-making under certainty, risk, & uncertainty. But how should organisms make decisions? Given incomplete information & limited cognitive resources, what constitutes rational choice? Normative models prescribe optimal decision-making strategies, while descriptive models describe how organisms actually make decisions.

Normative Models of Decision-Making: How We Should Decide

Normative models of decision-making are those models that address how humans should make decisions to be rational & maximize expected outcomes. These prescriptive models establish standards for optimal decision-making (Von Neumann & Morgenstern, 1944).

Normative models assume rational decision-makers who: (1) know all available options, (2) know all possible outcomes of each option, (3) can accurately estimate probabilities of outcomes, (4) can calculate expected values, & (5) consistently choose options with highest expected value. These assumptions are clearly unrealistic—real decision-makers face incomplete information, cognitive limitations, & inconsistent preferences. Nevertheless, normative models provide benchmarks against which actual decisions can be evaluated. They show what perfect rationality requires, even if organisms can’t achieve it.

Expected Utility Theory: Calculating Optimal Choices

Expected utility is calculated by multiplying the probability of each outcome times the utility (value) of that outcome. With multiple possible outcomes, this calculation is repeated for each outcome & the results are summed. The option with highest expected utility should be chosen (Von Neumann & Morgenstern, 1944).

Formula: Expected Utility = Σ [P(outcome) × U(outcome)], where P = probability & U = utility

Example: You’re offered a gamble. Pay $10 to flip a coin. If heads, win $30. If tails, win nothing. Should you accept? Calculate expected utility: Probability of heads = 0.5, Utility of winning $30 = $30; Probability of tails = 0.5, Utility of winning $0 = $0. Expected utility of gambling = (0.5 × $30) + (0.5 × $0) = $15. Cost of gambling = $10. Net expected utility = $15 – $10 = $5. Since expected utility is positive, expected utility theory prescribes accepting the gamble. Over many repetitions, you’d average a $5 gain per gamble. A rational decision-maker maximizing expected value should accept.

When decision-making under certainty, all probabilities equal 1.0 (100%), so expected utility reduces to simply comparing utilities of certain outcomes. The complexity of expected utility calculations becomes necessary only when probabilities are less than 1.0—that is, under conditions of risk or uncertainty.

Extensions: Stochastic Models of Choice

Stochastic models of choice (developed by R. Duncan Luce) treat preferences as though they have a random component. This allows us to understand how a person can prefer soup one day & salad the next, even when both options remain available & unchanged. Rather than assuming perfectly consistent preferences, stochastic models acknowledge variability in choice. A person might prefer soup 70% of the time & salad 30% of the time. This probabilistic approach better captures actual human behavior than deterministic models assuming perfect consistency (Luce, 1959).

Subjective Utility Theory: When Probabilities Are Unknown

Subjective utility theory is an extension of expected utility theory which is used when a probability of an outcome or event cannot be determined in advance—that is, under conditions of uncertainty rather than risk. Decision-makers must estimate subjective probabilities based on beliefs, experience, & incomplete information (Savage, 1954).

When facing uncertainty, you don’t know objective probabilities but can estimate subjective probabilities—your personal assessment of likelihood based on available information. Subjective utility theory proposes that rational decision-makers under uncertainty should: (1) assign subjective probabilities to outcomes, (2) calculate expected utilities using these subjective probabilities, & (3) choose the option with highest subjective expected utility.

Decision Trees: Visualizing Complex Choices

A decision tree is a visual tool that might be used to assist someone—for instance, a jury member deciding whether to convict or acquit when not absolutely certain of a defendant’s guilt. Decision trees map out all possible choices, outcomes, & their consequences, helping decision-makers systematically evaluate options.

Example decision tree for a juror: Choice 1: Vote to Convict—Branch 1a: Defendant is actually guilty (probability = 0.8 based on evidence), Outcome: Justice served, guilty person punished (high positive utility); Branch 1b: Defendant is actually innocent (probability = 0.2), Outcome: Innocent person imprisoned (very high negative utility). Choice 2: Vote to Acquit—Branch 2a: Defendant is actually guilty (probability = 0.8), Outcome: Guilty person goes free (moderate negative utility); Branch 2b: Defendant is actually innocent (probability = 0.2), Outcome: Innocent person freed (high positive utility).

The juror must weigh: Is avoiding false conviction of an innocent person more important than ensuring conviction of a guilty person? The “beyond reasonable doubt” standard reflects society’s judgment that false convictions are worse than false acquittals, setting a high threshold for conviction.

Decision trees help in countless contexts: Should you go to college? (branches for attend/don’t attend, succeed/fail, career outcomes). Should you accept a job offer? (branches for accept/decline, job works out/doesn’t, alternative opportunities). By making choices, probabilities, & outcomes explicit, decision trees reduce cognitive load & reveal trade-offs.

Utilitarianism: Considering Effects on Others

Utilitarianism is a theory of morality that argues that the best choice is always the one which maximizes the utility of all people that will be affected by the choice. Rather than maximizing personal utility, one should maximize total utility across all affected individuals (Mill, 1863).

Utilitarianism extends decision theory from individual to collective outcomes. Should we only consider consequences to ourselves when making choices? Shouldn’t we also consider how our choices affect others? Utilitarian ethics says yes—moral choices maximize aggregate happiness or well-being, even if this requires personal sacrifice.

Example: If a doctor has an otherwise healthy patient come in for an appendix removal, is the doctor justified in harvesting the organs (heart & two kidneys) of the appendicitis patient & using these organs to save the lives of three patients waiting for transplants? From a strict utilitarian calculus: one death produces three lives saved, net utility increases. Most people recoil from this conclusion, suggesting pure utilitarianism conflicts with moral intuitions about individual rights, consent, & the difference between killing & letting die.

This extreme example reveals limitations of pure utilitarianism. While considering effects on others is ethically important, maximizing aggregate utility without constraints can justify horrific actions. Modified utilitarianism incorporates side constraints—rights & principles that cannot be violated even to maximize utility. Nevertheless, utilitarian thinking pervades policy decisions: public health interventions, resource allocation, environmental regulations, & welfare policies often aim to maximize societal well-being, accepting costs to some individuals for greater collective benefit.

Probability Theory: Optimal Predictions in Uncertain Worlds

Probability theory is a branch of mathematics that deals with random events & allows for the best “average” prediction to be made. Most famously used by gamblers, probability theory provides formal methods for reasoning about uncertain outcomes (Kolmogorov, 1933).

Probability theory establishes mathematical foundations for reasoning under uncertainty. It defines how to calculate probabilities, combine probabilities (multiplication rule, addition rule), & update probabilities based on new evidence (Bayes’ theorem). These tools enable rational inference from incomplete information.

However, applying probability theory requires understanding relationships between different types of errors—particularly in contexts like medical diagnosis, criminal justice, & statistical hypothesis testing. Two types of errors exist:

Type I Error (False Positive): Concluding something is true when it’s actually false. Examples: Convicting an innocent person, diagnosing disease in a healthy person, concluding a drug works when it doesn’t.

Type II Error (False Negative): Concluding something is false when it’s actually true. Examples: Acquitting a guilty person, failing to diagnose disease in a sick person, concluding a drug doesn’t work when it actually does.

These errors trade off: Making one type less likely makes the other type more likely. Lowering the conviction threshold reduces false acquittals (Type II errors) but increases false convictions (Type I errors). Raising the threshold does the opposite. “Beyond reasonable doubt” sets a high threshold, prioritizing avoidance of false convictions at the cost of more false acquittals. Different contexts weight these errors differently based on their consequences.

Descriptive Models of Decision-Making: How We Actually Decide

Descriptive models of decision-making are those models that address how humans actually make decisions, describing observed behavior rather than prescribing optimal behavior. These models reveal systematic deviations from normative prescriptions (Kahneman & Tversky, 1979).

Bounded Rationality: The Limits of Human Reason

Herbert Simon introduced the concept of bounded rationality to explain why humans don’t—and can’t—make decisions the way normative models prescribe. Bounded rationality recognizes that human rationality is limited by cognitive constraints: finite information-processing capacity, limited working memory, restricted attention, & time pressures (Simon, 1956).

Normative models assume decision-makers can: (1) identify all available options, (2) determine all possible outcomes, (3) accurately assess probabilities, (4) calculate expected utilities, & (5) compare all options simultaneously. These assumptions require computational abilities humans simply don’t possess. Real decision-makers face information overload—too many options, too much data, too little time. They can’t hold all relevant information in mind simultaneously or perform complex probability calculations mentally.

Bounded rationality doesn’t mean humans are irrational—it means human rationality operates within constraints. Given these constraints, satisficing & heuristic use are rational adaptations, not failures. A decision-maker who satisfices isn’t being lazy or irrational; they’re appropriately allocating limited cognitive resources. Spending hours optimizing a trivial decision wastes resources better devoted to important choices.

Simon argued that understanding actual decision-making requires studying the interaction between the mind’s limitations & the environment’s structure. Some environments are “friendly” to bounded rationality—they provide clear feedback, allow learning from mistakes, & don’t punish satisficing severely. Other environments are hostile—they provide misleading cues, delayed feedback, & severe penalties for suboptimal choices. Effective decision-making involves matching strategies to environmental demands while respecting cognitive limitations.

Normative models establish rational benchmarks. Descriptive models document reality. The gap between them is substantial: humans routinely violate expected utility theory, make inconsistent choices, ignore probabilities, & rely on cognitive shortcuts producing systematic errors. Understanding actual decision-making requires descriptive models capturing these patterns.

Satisficing: Good Enough Is Good Enough

Proposed by Herbert Simon, a Nobel Prize winner in economics, satisficing represents a fundamental departure from optimization. People “satisfice” rather than optimize when they make decisions.

Satisficing is a theory of decision-making that argues that in the face of uncertainty & incomplete information, decision-makers work to satisfy basic needs & achieve acceptable (but not necessarily optimal) outcomes rather than searching exhaustively for the absolute best option (Simon, 1956).

To satisfice is to satisfy your most important needs, even though there may be other options (undiscovered or not fully considered) that might satisfy your needs better. A satisficer sets an acceptability threshold—a level of outcome quality that’s “good enough.” They evaluate options sequentially, selecting the first option exceeding this threshold rather than exhaustively comparing all options to find the absolute best.

Example: Buying a used car. An optimizer would research every available car in their price range, test drive dozens of vehicles, analyze reliability data, calculate total cost of ownership, & select the objectively best car. A satisficer determines acceptable criteria (reliable, fuel-efficient, under $15,000, drives smoothly), evaluates cars sequentially, & purchases the first car meeting these criteria. They accept “good enough” rather than pursuing “best possible.”

Why Do Humans Satisfice?

Satisficing is more efficient. Exhaustive search costs time, effort, & cognitive resources. In many contexts, the improvement from finding the absolute best option doesn’t justify the search costs.

Often there is no penalty for guessing wrong. If you choose a decent restaurant rather than the optimal restaurant, the consequences are minimal. Life continues. The stakes don’t justify extensive deliberation.

Weighing different options may not improve your chances of guessing correctly anyway. With incomplete information & unpredictable outcomes, extensive analysis may not actually increase success probability. Sometimes quick decisions based on satisficing heuristics perform as well as laborious optimization attempts.

Satisficing Versus Maximizing: Who’s Happier?

So who’s happier, satisficers or maximizers? Research by Schwartz et al. (2002) found that satisficers report greater life satisfaction, happiness, optimism, & self-esteem than maximizers. Maximizers experience more regret, depression, & perfectionism. Even when maximizers achieve objectively better outcomes (higher salaries, better grades), they’re less satisfied because they always wonder if an even better option exists. Satisficers accept “good enough,” feel content with choices meeting their standards, & move on without endless second-guessing. The psychological costs of maximizing often outweigh its objective benefits.

Personality & Decision Style: Type A Versus Type B

Individual differences in personality influence whether people tend toward maximizing or satisficing strategies. Psychologists have long distinguished between Type A & Type B personality patterns, originally identified in research on cardiovascular health (Friedman & Rosenman, 1974).

Type A personality is characterized by competitiveness, time urgency, hostility, & achievement striving. Type A individuals are driven, impatient, & focused on accomplishing as much as possible in minimal time. They set high standards, work intensely, & often feel pressure to outperform others. In decision-making contexts, Type A individuals tend toward maximizing—exhaustively searching for the best possible option, comparing alternatives extensively, & feeling dissatisfied unless they’ve found the optimal choice.

Type B personality is characterized by a more relaxed, patient, & easygoing approach to life. Type B individuals are less driven by competition & time pressure. They work steadily but don’t feel compelled to accomplish everything immediately. In decision-making contexts, Type B individuals tend toward satisficing—accepting “good enough” options that meet their criteria without exhaustively searching for the absolute best. They’re more comfortable with uncertainty about whether better options exist.

This personality-decision style connection has important implications. Type A maximizers may achieve objectively better outcomes in some contexts—finding better deals, securing higher salaries, selecting superior products. However, as Schwartz et al. (2002) demonstrated, maximizers also experience more regret, anxiety, & dissatisfaction. Type B satisficers may miss some optimal choices but enjoy greater contentment with their decisions. The “best” decision strategy depends partly on personality fit—strategies that work well for Type B individuals may frustrate Type A individuals, & vice versa.

Understanding your own personality tendencies can improve decision-making. Type A individuals might benefit from deliberately practicing satisficing in low-stakes decisions, conserving cognitive resources for truly important choices. Type B individuals might benefit from occasionally pushing toward more thorough search in high-stakes decisions where finding the best option matters significantly. Matching decision strategy to both the situation & one’s personality produces better outcomes than rigidly applying one approach to all decisions.

Prospect Theory: How Gains & Losses Differ

Prospect theory is a descriptive theory of choice under risk developed to account for differences between the ideal perfect choices predicted by expected utility theory & the actual choices people make. It demonstrates that people evaluate options relative to reference points & weight losses more heavily than equivalent gains (Kahneman & Tversky, 1979).

Prospect theory’s key insight: The value function for gains is different from the value function for losses. The value function for gains tends to be somewhat shallow whereas the value function for losses is quite steep. This means that a loss of $1000 is more “painful” than the gain of $1000 is “joyful.” This asymmetry leads to loss aversion in decision-making—people are more motivated to avoid losses than to acquire equivalent gains.

Loss aversion produces several phenomena:

Endowment Effect: Once you own something, you value it more highly than before you owned it. People demand more money to sell an item they own than they would pay to acquire that same item. The reference point shifts—losing the item feels like a loss from the new reference point (ownership), so it must be compensated more heavily.

Status Quo Bias: People prefer maintaining current situations over changing to new situations, even when the new situation is objectively better. Change feels like a loss from the current reference point, so it requires substantial benefit to overcome loss aversion.

Risk Aversion for Gains, Risk Seeking for Losses: People avoid risk when choosing between gains (prefer certain $50 over 50% chance of $100) but seek risk when choosing between losses (prefer 50% chance of losing $100 over certain loss of $50). This violates expected utility theory’s assumption that risk preferences should be consistent.

The Reflection Effect: Mirror-Image Risk Preferences

The reflection effect describes how risk preferences reverse—or reflect—between the domains of gains & losses. When outcomes are framed as gains, people are risk averse (preferring certainty). When the same outcomes are framed as losses, people become risk seeking (preferring gambles). This reflection produces mirror-image patterns across the gain-loss boundary.

Example: Given a choice between a certain gain of $50 or a 50% chance of gaining $100, most people choose the certain $50 (risk aversion for gains). But given a choice between a certain loss of $50 or a 50% chance of losing $100, most people choose the gamble (risk seeking for losses). The expected values are identical in both cases ($50), yet preferences reverse depending on whether outcomes are gains or losses. This violates expected utility theory, which assumes consistent risk preferences regardless of domain.

The reflection effect has profound practical implications. Investors hold losing stocks too long (hoping to avoid realizing losses) while selling winning stocks too quickly (locking in certain gains). Negotiators make different concessions depending on whether they frame outcomes as gains from a starting point or losses from an aspiration point. Medical patients choose differently when treatments are described in terms of survival rates versus mortality rates.

The Certainty Effect: Overweighting Sure Things

The certainty effect describes people’s tendency to overweight outcomes that are certain relative to outcomes that are merely probable. Reducing probability from 100% to 95% has much greater psychological impact than reducing probability from 50% to 45%, even though the mathematical difference is identical (5 percentage points).

Example: Most people prefer a certain $30 over an 80% chance of $45 (expected value = $36). The certainty of $30 outweighs the higher expected value of the gamble. But when both options are made uncertain—25% chance of $30 versus 20% chance of $45—preferences often reverse toward the higher expected value option. Certainty has special psychological weight that mere probability lacks.

The certainty effect explains insurance purchasing (paying premiums to convert uncertain losses into certain small costs), preference for guaranteed salary over commission-based pay, & reluctance to trade certain benefits for larger but uncertain ones. It also explains the pseudocertainty effect—treating outcomes as certain when they’re actually conditional on uncertain events. If told “you have a 25% chance of reaching the final round, where you’ll definitely win $30,” people treat the $30 as more certain than it actually is (only 25% probable), because within the conditional scenario, it’s guaranteed.

Prospect theory revolutionized understanding of decision-making under risk, explaining systematic violations of expected utility theory. It earned Daniel Kahneman the Nobel Prize in Economics (Tversky had passed away) & became foundational to behavioral economics.

Regret Theory: Avoiding the Pain of Wrong Choices

Regret theory is a descriptive model of decision-making that assumes that decision-makers experience feelings of regret & try to avoid this feeling of regret when making decisions. People anticipate potential regret & choose options minimizing expected regret (Loomes & Sugden, 1982).

Regret is the painful feeling experienced when realizing a different choice would have produced a better outcome. Anticipated regret influences decisions: People avoid options with high regret potential even when expected utility is favorable. They also take actions specifically to reduce potential regret, such as gathering more information before deciding (to avoid regret about deciding hastily) or maintaining flexibility (to avoid regret about commitment).

Example: You’re choosing between two job offers. Job A pays $10,000 more annually but requires relocating. Job B allows staying in your current city. You choose Job A, relocate, & later learn Job B employees received a huge bonus & profit-sharing plan, making their total compensation exceed Job A. You experience intense regret: “If only I’d chosen Job B!” Anticipated regret about moving away from family & friends might have led you to choose Job B despite lower initial salary, specifically to avoid this potential regret.

The Sunk Cost Fallacy: Throwing Good Money After Bad

The sunk cost fallacy is the tendency to continue investing in something because of previously invested resources (time, money, effort) that cannot be recovered. Rational decision-making should consider only future costs & benefits—past investments are “sunk” & irrelevant to optimal future choices. Yet people routinely let sunk costs influence decisions, often to their detriment (Arkes & Blumer, 1985).

Example: You’ve paid $100 for a concert ticket, but on the night of the concert, you feel ill & would rather stay home. Rationally, the $100 is gone regardless of whether you attend—it’s a sunk cost. The decision should depend only on whether attending (while ill) or staying home provides greater utility tonight. But many people force themselves to attend, reasoning “I can’t waste that $100!” The sunk cost irrationally influences behavior.

The sunk cost fallacy appears across domains: Investors hold losing stocks hoping to “break even” rather than selling & reinvesting in better opportunities. Companies continue failing projects because they’ve already invested millions. Relationships persist past their expiration because of “all the years we’ve put in.” Students finish degrees they’ve lost interest in because of completed coursework. In each case, past investments—which cannot be recovered—irrationally influence future decisions.

Several psychological mechanisms drive the sunk cost fallacy. Loss aversion makes abandoning investments feel like accepting a loss, which is psychologically painful. Regret avoidance makes people fear the regret of admitting a bad investment. Self-justification motivates continued investment to prove the original decision wasn’t wrong. Waste aversion creates discomfort with “wasted” resources, even when continued investment wastes even more.

Overcoming the sunk cost fallacy requires deliberately ignoring past investments when evaluating options. Ask: “If I were starting fresh today, with no prior investment, which option would I choose?” This “zero-based” thinking focuses attention on future consequences rather than past costs. Organizations can institutionalize this by requiring periodic project reviews that evaluate continuation based solely on projected future returns, not historical investment.

Compensatory Decision-Making Strategies

Compensatory strategies involve trading low values on one dimension for high values on another. A car buyer might accept poor gas mileage (low value on efficiency dimension) in exchange for exceptional reliability (high value on dependability dimension) (Payne, Bettman, & Johnson, 1993).

Compensatory strategies allow trade-offs: Weaknesses on some dimensions can be offset by strengths on others. Total evaluation integrates across dimensions. This contrasts with non-compensatory strategies where deficiency on any critical dimension eliminates an option regardless of strengths. Compensatory strategies better approximate expected utility calculations, weighing multiple factors & integrating information. However, they’re cognitively demanding, requiring dimensional comparisons & trade-off calculations. Under time pressure or cognitive load, decision-makers often shift to simpler non-compensatory strategies (eliminate options failing to meet minimum thresholds on any dimension).

Looking Forward

We’ve distinguished normative models prescribing optimal decision-making from descriptive models describing actual behavior. Expected utility theory provides a rational benchmark. Satisficing replaces optimization with “good enough” standards. Prospect theory demonstrates loss aversion: losses loom larger than equivalent gains. Regret theory shows how anticipated regret shapes choices. In Part 4, we’ll explore intuition, heuristics, & biases—mental shortcuts that enable rapid decision-making but produce systematic errors, including the availability heuristic, representativeness heuristic, anchoring, framing, & overconfidence.