Week 08
Consumer Behavior · Davood Wadi, PhD
Consumers follow predictable mental shortcuts. In some conditions, so do the AI agents that shop for them.
Classical economics assumes that consumers compare every option and choose the one with the highest value.
Behavioral economics studies how people actually make economic decisions.
Herbert Simon proposed that people have bounded rationality. They have limited time, information, and attention.
Because of these limits, people often choose an option that is good enough instead of the best possible option. Simon called this satisficing.
The departures from the rational model are not random. They follow patterns that researchers can predict and marketers can use.
Psychologists describe two modes of thinking (Kahneman, 2011).
System 1 is fast, automatic, and effortless. It recognizes a familiar logo or reacts to a red "SALE" sign in an instant.
System 2 is slow, deliberate, and effortful. It compares the price per unit of two packages or reads the terms of a phone contract.
Most everyday purchases rely on System 1.
System 2 takes over when the decision is important, unfamiliar, or clearly wrong at first sight, and when the consumer has the time and energy to think.
Each card shows a consumer thinking about a purchase. Is the consumer using fast, automatic System 1 or slow, deliberate System 2?
Tap a card, then tap the group it belongs to.
A heuristic is a mental shortcut that simplifies a decision.
The availability heuristic judges how likely something is by how easily examples come to mind. A consumer who recently heard about a car recall may overestimate the risk of that brand.
The representativeness heuristic judges an item by how closely it resembles a familiar category. A store brand in packaging like the market leader seems to share its quality.
The price-quality inference assumes that a higher price signals higher quality. Consumers rely on it most when they cannot judge quality directly, as with wine or perfume.
Heuristics save time and often work well. They lead to errors when the shortcut does not fit the situation.
Each consumer takes a mental shortcut. Which heuristic is it?
Availability: recent examples come easily to mind, so Dev judges the risk as higher.
Price-quality inference: when quality cannot be judged directly, a higher price is read as higher quality.
Representativeness: the can resembles the market leader, so it seems to share its quality.
Availability: one vivid example that comes easily to mind makes the event seem common.
Price-quality inference: the higher price is taken as a signal of quality.
Representativeness: the box resembles a familiar category, so Ada judges it as belonging to it.
Which concept does this case show?
Anchoring occurs when an initial number influences a later judgment, even when the number is irrelevant.
A "compare at $120" label makes a $79 price seem lower than it would appear on its own.
In one classic study, people who first wrote down the last two digits of their social security number were then willing to pay more for products when those digits were high (Ariely, Loewenstein, & Prelec, 2003).
Marketers use anchors in reference prices, suggested quantities such as "limit 12 per customer," and high-priced options placed first on a menu.
A furniture store places a $2,400 sofa at the entrance, beside the $1,100 sofa it most wants to sell. Which explanation fits what the slide describes?
Prospect theory states that people judge outcomes as gains or losses relative to a reference point (Kahneman & Tversky, 1979).
Loss aversion means that a loss feels about twice as strong as a gain of the same size.
Losing $20 hurts more than finding $20 pleases.
The endowment effect means that people value an item more once they own it.
Free trials and money-back guarantees use this effect. After a consumer has used a product for a month, giving it up feels like a loss.
Each card describes a consumer's reaction. Does it show loss aversion or the endowment effect?
Tap a card, then tap the group it belongs to.
Status quo bias is the preference for keeping things as they are. Consumers stay with the same bank, phone plan, or streaming service even when a better offer exists.
A framing effect occurs when the way information is presented changes the choice, even though the facts are the same. Consumers rate ground beef more favorably when it is described as "75% lean" than as "25% fat" (Levin & Gaeth, 1988).
A surcharge for paying by card and a discount for paying in cash can describe the same price gap, but consumers react more strongly to the surcharge because it is framed as a loss.
Each card describes a consumer's choice. Is it status quo bias or a framing effect?
Tap a card, then tap the group it belongs to.
Just-below pricing ends a price just under a round number, such as $4.99 instead of $5.00.
Consumers read prices from left to right and give extra weight to the leftmost digit. This is called the left-digit effect (Thomas & Morwitz, 2005).
A price of $4.99 therefore feels closer to $4 than to $5.
Promotional framing presents a price as a discount from a higher list price, such as "20% off."
The discount adds a feeling of a good deal, called transaction utility, on top of the value of the product itself (Thaler, 1985).
Both cues work through System 1. A consumer who calculates the price per unit is less affected by them.
Choice architecture is the way options are arranged and presented to the decision maker (Thaler & Sunstein, 2008).
A default is the option a person receives if they do nothing. Many people keep the default because changing it takes effort.
In countries where organ donation is the default, far more people are registered donors than in countries where people must opt in (Johnson & Goldstein, 2003).
A nudge is a change in choice architecture that steers behavior without removing any option or changing its price. Placing fruit at eye level in a cafeteria is a nudge.
A nudge is transparent and serves the person's own interest. A dark pattern uses the same tools against the person's interest.
Each case describes how options are arranged for a consumer. Is it a default, a nudge, or a dark pattern?
Default: the option a person receives if they do nothing.
Nudge: the arrangement steers behavior without removing any option or changing its price, in the diner's own interest.
Dark pattern: the design works against the consumer's interest.
Default: customers who do nothing receive the renewable plan.
Dark pattern: the same tool as a default, used against the person's interest and not transparent.
Nudge: transparent, it serves the person's own interest, and no option is removed.
Which concept does this case show?
AI agents learn from text written by people, so researchers ask whether agents also show human biases.
Some studies find heuristic-like choices. AI models stated a higher willingness to pay for a product after seeing a high number in the question, which is an anchoring effect (Wadi & Fredette, 2025).
In that study, newer models were more susceptible to anchoring than older models. A newer model is not always a less biased one (Wadi & Fredette, 2025).
AI agents can also favor options because of where they appear in a list, which is a position effect (Wadi & Ma, 2026a).
Other studies find rational choices. When all product details were visible, seven of eight AI agents gave the cents in a price the same weight as whole dollars, so a $4.99 price did not mislead them (Wadi & Ma, 2026b).
Likewise, six of eight agents gave a dollar saved through a discount the same weight as a dollar of list price (Wadi & Ma, 2026b).
Whether an agent shows a bias therefore depends on the conditions in which it chooses.
A company upgrades its shopping agent to the newest AI model, expecting it to be immune to pricing tricks. Based on the studies on the slide, what should the company expect?
Storefront information architecture is the way a store organizes product information and how easy it is to reach.
A human shopper who must click into each product page to find its weight faces a search cost. An AI agent faces the same kind of cost in time, computing, or money.
When the details were hidden behind costly look-ups and the goal was vague, AI agents often skipped the details needed to compare value (Wadi & Ma, 2026b).
Their choices then looked like human heuristics. They chose the lower whole-dollar price or the lower list price, even when it was the worse deal (Wadi & Ma, 2026b).
When the agents were given the missing details, most chose the better product, and none made an error when asked to calculate the price per unit (Wadi & Ma, 2026b).
In most cases, the heuristic-like choice came from the information the agent held, not from an error in its reasoning.
Retailers who make diagnostic details, such as weight and unit price, easy for agents to reach help agents choose well. Rules that require unit prices on shelf labels were written for human shoppers and do not yet cover what an agent can reach.
An online grocer finds that AI agents shopping on its site often pick the bag of rice with the lower price even when it costs more per kilogram. Each product's weight appears only on its product page. Which change fits what the slide describes?
Find an online product page or a store shelf. Which heuristics, anchors, or frames does it use? Would an AI agent shopping for you be affected by the same cues, and which details would it need to see to avoid them?