# Psychology experiments for students to run live in class Source: https://decka.dev/ideas/psychology-experiments-for-students Psychology experiments for students to run live in class. Framing, anchoring, the decoy effect, the Linda problem, Wason's cards, and Asch and Milgram as predictions. Some famous experiments started in a classroom. Tversky and Kahneman's framing study took its answers from students at Stanford and the University of British Columbia who [answered brief questionnaires in a classroom setting](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf). Run one again with your own class and the result stops being a line in a book. It becomes something the room did a minute ago, on its own phones. Here are ten that work live. Four split the room in half and give each half a different version of the same question. Four catch the whole room in a reasoning trap at once. Two, Asch and Milgram, are not things to stage on students, so the room predicts the result and you show what the original found. Each section says what to ask, what the original study found, and why it happens. ## How to run a split-room experiment with a class Several well-known results in social psychology come from giving two groups the same choice in different words. Tversky and Kahneman's framing study did it with [paper questionnaires](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf). On phones, each student is put in one group or the other at random and sees only their own version. **Do not say there are two versions until the reveal.** Each half should answer its own question as if it were the only one. **Let everyone answer before anything goes on the screen**, so no answer is given with the room's results in view. **Show the two versions side by side, then the answers**, so the room can check that the questions describe the same thing before it compares its answers. Decka's [A/B Tests](/ab-testing) activity works this way. Each phone gets one version at random, the big screen puts the two versions side by side, and "Reveal results" shows both halves' answers on one chart. ## Framing effect example: the disease problem **What you ask.** In Tversky and Kahneman's version, a disease is expected to kill [600 people](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf), and two programs have been proposed. Half the room reads them as lives saved. One program saves [200 people](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) for certain. The other has a [one-in-three chance](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) of saving [all 600](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) and a [two-in-three chance](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) of saving nobody. The other half reads the same two programs as deaths. Either [400 people](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) die for certain, or there is a [one-in-three chance](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) that nobody dies and a [two-in-three chance](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) that [all 600](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) die. **What the original found.** Tversky and Kahneman gave the two versions to two groups of students. Described as lives saved, [72% of 152](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) chose the certain program. Described as deaths, [22% of 155](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) did. They saw the same reversal in [university faculty and physicians](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf). **Why it happens.** In the authors' words, [choices involving gains are often risk averse and choices involving losses are often risk taking](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf). The wording decides which of the two the same outcome feels like. **Whether it holds up.** The Many Labs project reran thirteen classic effects across [36 samples](https://osf.io/wx7ck/), and this one held, at about [half its original size](https://osf.io/wx7ck/). Its version dropped the word "Asian" from the original's "unusual Asian disease", and so should yours. ## Anchoring bias experiment: a number nobody should trust **What you ask.** Half the room sees a low number and the other half a high one, and then everyone estimates the same quantity. Tversky and Kahneman spun a wheel of fortune in front of their subjects. They asked whether the [percentage of African countries in the United Nations](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf) was higher or lower than the number it landed on, and then for an estimate. **What the original found.** Groups that started from [10 and 65](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf) gave median estimates of [25 and 45](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). Paying for accurate answers did not shrink the gap. **Why it happens.** People start from the number in front of them and adjust away from it, and [adjustments are typically insufficient](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). In the original, [a number the subjects watched a wheel of fortune produce](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf) still pulled their estimates towards it. **Whether it holds up.** Many Labs used different questions, such as the height of Mount Everest, with an anchor that was clearly too high or too low. Anchoring turned up in [every sample](https://osf.io/wx7ck/), on all four questions. ## Decoy effect example: the option nobody picks **What you ask.** Half the room chooses between two options, each better on one thing: a cheaper one and a better one. The other half gets a third option as well, plainly worse than one of the two in every way. Nobody should choose it, and [almost nobody does](https://people.duke.edu/~jch8/bio/Papers/HuberPaynePutoJMR%202014.pdf). **What the original found.** Joel Huber, John Payne and Christopher Puto showed in 1982 that adding such a decoy can raise the share of the option it sits next to. That breaks a basic rule of rational choice, that [one cannot increase the probability of choosing an alternative by adding another alternative to the choice set](https://people.duke.edu/~jch8/bio/Papers/HuberPaynePutoJMR%202014.pdf). In their studies the decoy itself took about [2%](https://people.duke.edu/~jch8/bio/Papers/HuberPaynePutoJMR%202014.pdf) of choices. **Whether it holds up.** The authors are candid that it is not universal. Their [first experiments showed positive results for some of the choice sets, but not for others](https://people.duke.edu/~jch8/bio/Papers/HuberPaynePutoJMR%202014.pdf), and they say the effect [largely replicates](https://people.duke.edu/~jch8/bio/Papers/HuberPaynePutoJMR%202014.pdf) when it is tested the classic way. Present it as a case of context changing a choice, not as a law. ## Question wording: forbid or allow **What you ask.** Half the room is asked whether the government should forbid public speeches against democracy. The other half is asked whether it should allow them. Both halves answer yes or no. **What the original found.** In Rugg's original, [62%](https://osf.io/wx7ck/) said such speeches should not be allowed, but only [46%](https://osf.io/wx7ck/) said they should be forbidden. Many Labs found the gap again in [97% of its samples](https://osf.io/wx7ck/). **One way to read it.** "Forbid" may sound harsher than "not allow", though the two policies are the same, so ask the room whether it does. Either way, the room's against answers sit on opposite buttons: yes to forbid, no to allow. **The famous version needs a film.** Elizabeth Loftus and John Palmer showed films of traffic accidents to [45 students](https://gwern.net/doc/psychology/cognitive-bias/1974-loftus.pdf) and asked how fast the cars were going. Asked with "smashed", they estimated [40.8](https://gwern.net/doc/psychology/cognitive-bias/1974-loftus.pdf) miles an hour on average; asked with "contacted", [31.8](https://gwern.net/doc/psychology/cognitive-bias/1974-loftus.pdf). A second experiment used new students. A week after its film, [32%](https://gwern.net/doc/psychology/cognitive-bias/1974-loftus.pdf) of those asked with "smashed" remembered broken glass that was not there, against [14%](https://gwern.net/doc/psychology/cognitive-bias/1974-loftus.pdf) of those asked with "hit". The question changes a memory of something the room saw, so it is not in the deck. To try it, play a short clip of a collision first and ask the speed in the two wordings. ## Availability heuristic example: the letter R **What you ask.** Whether the letter R turns up more often as the first letter of English words or as the third. Everyone answers alone, at the same moment. **What the original found.** Tversky and Kahneman used this very letter. In their account, [most people judge words starting with a consonant to be more numerous](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf) than words with it third. They do so [even for consonants, such as r or k, that are more frequent in the third position than in the first](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). **Why it happens.** People answer by [recalling words that begin with r (road) and words that have r in the third position (car)](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf), and go by which come to mind more easily. [It is much easier to search for words by their first letter than by their third letter](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). Red, river and run arrive at once. Word, care and park take effort, so they feel rarer. **A caution.** The letter result has been challenged. Sedlmeier, Hertwig and Gigerenzer ran three studies in which judgments of whether a letter is commoner in first or second position [generally followed the actual proportions](https://edoc.unibas.ch/8242/). Teach availability with it, and tell the class the letter version is disputed. ## The conjunction fallacy and the Linda problem **What you ask.** Read Linda's description aloud, then ask which is more probable: that she is a bank teller, or that she is a bank teller and active in the feminist movement. Word for word, [Tversky and Kahneman's description](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf) is this: "Linda is 31 years old, single, outspoken and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations." **What the original found.** Tversky and Kahneman put exactly that choice to [142 undergraduates](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf) at the University of British Columbia, and [85%](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf) picked the second. It cannot be the more probable: [the probability of a conjunction, P(A&B), cannot exceed the probabilities of its constituents](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf). **Why it happens.** [A conjunction can be more representative than one of its constituents](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf). Linda sounds like an activist, so the longer description feels like the truer one. **What to discuss.** The authors did not believe it either. They [searched for alternative interpretations](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf), then ran what they called [a series of increasingly desperate manipulations](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf) to get students to obey the rule. Ask your class what they meant by "probable". **For a whole lesson.** Decka's [conjunction fallacy lesson](/classroom/psychology/conjunction-fallacy) runs the Linda problem as a full class: each student answers alone, argues with someone who chose the other, then answers again. ## Gambler's fallacy example: heads, heads, heads **What you ask.** Which of two runs of six coin tosses is more likely: heads, tails, heads, tails, tails, heads, or three heads followed by three tails. Offer "equally likely" as a third answer. **What the original found.** Tversky and Kahneman report that [people regard the sequence H-T-H-T-T-H to be more likely than the sequence H-H-H-T-T-T](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). They are equally likely. So is any other run of six. **Why it happens.** [Chance is commonly viewed as a self-correcting process](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf), so a run that looks lopsided looks unlikely. In fact, [deviations are not "corrected" as a chance process unfolds, they are merely diluted](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). A follow-up question works well here: after five heads in a row, is tails due? ## The Wason selection task: four cards **What you ask.** Four cards each have a letter on one side and a number on the other. You can see E, K, 4 and 7. The rule is that a card with a vowel on one side has an even number on the other. Which cards must you turn over to find out whether the rule is true? **The answer.** E and 7. An odd number behind the E breaks the rule, and so does a vowel behind the 7. The 4 cannot break it, whatever is on its back. **What the original found.** Peter Wason's abstract version beat nearly everyone. Across four experiments, [5 of 128](https://iphils.uj.edu.pl/~a.trybus/classes/logic/cs/15.Wason.Natural.And.Contrived.Reasoning.pdf) students chose the right pair, and the commonest choice was the vowel and the even number. Wason and Shapiro describe those students as [dominated by verification rather than falsification](https://iphils.uj.edu.pl/~a.trybus/classes/logic/cs/15.Wason.Natural.And.Contrived.Reasoning.pdf): they looked for cards that could confirm the rule. **Why it matters.** The same logic is easier in a familiar setting. With a rule about travel ("Every time I go to Manchester I travel by car"), [10 of 16](https://iphils.uj.edu.pl/~a.trybus/classes/logic/cs/15.Wason.Natural.And.Contrived.Reasoning.pdf) students solved it, against [2 of 16](https://iphils.uj.edu.pl/~a.trybus/classes/logic/cs/15.Wason.Natural.And.Contrived.Reasoning.pdf) on the abstract cards. ## The Asch conformity experiment, as a prediction **Why the room predicts it.** Asch needed [a room of helpers who gave the wrong answer on purpose](https://archive.org/details/solomon-asch_confirmaty), and one person who did not know. In a class that means deceiving one student in front of their peers. Ask the room to predict the result instead. **What you ask.** Asch's subjects matched the length of a line to one of three others, [which almost nobody gets wrong alone](https://archive.org/details/solomon-asch_confirmaty). When the people around them all gave the same wrong answer first, how often did they go along? **What the original found.** Asch tested [123 subjects](https://archive.org/details/solomon-asch_confirmaty) from three colleges. In ordinary circumstances people matching the lines [make mistakes less than 1 per cent of the time](https://archive.org/details/solomon-asch_confirmaty). Under group pressure they went with the wrong majority in [36.8 per cent of the selections](https://archive.org/details/solomon-asch_confirmaty). [About one quarter of the subjects](https://archive.org/details/solomon-asch_confirmaty) never agreed with the majority at all. **What to discuss.** [All the yielding subjects underestimated the frequency with which they conformed](https://archive.org/details/solomon-asch_confirmaty). Ask the room how often it thinks it would have gone along. The article first appeared in [Scientific American in November 1955](https://www.scientificamerican.com/article/opinions-and-social-pressure/). ## The Milgram experiment, as a prediction **Why the room predicts it.** Milgram told volunteers to give a man in another room stronger and stronger electric shocks. No shock was real, and the volunteers did not know that. Milgram described what it did to them: [Profuse sweating, trembling, and stuttering were typical expressions of this emotional disturbance](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf). Nobody should recreate that in a classroom, even as a joke. **What you ask.** Of the [40 men](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf) in Milgram's study, what share went all the way to the last switch on the board, marked [450 volts](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf)? **What the original found.** [26 of the 40](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf) obeyed to the end. Beforehand, [fourteen Yale seniors](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf), all psychology majors, predicted on average that [1.2%](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf) would. **What to discuss.** The gap between the prediction and the result is the lesson. Ask whether your room's guesses landed nearer the seniors' [1.2%](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf) or the [26 of 40](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf) who went to the end. Then show that the reading of the result is still argued over. Haslam and Reicher argue that [individuals' willingness to follow authorities is conditional on identification with the authority in question](https://doi.org/10.1371/journal.pbio.1001426). ## Three more psychology activities already built for a class - **[The Stroop effect](/classroom/psychology/stroop-effect).** Every student times four colour-word cards on their phone, then sees how much the words slowed the whole room. - **[An optical illusions game](/classroom/psychology/optical-illusions-game).** Every phone matches two lines by eye, and the room's matches land on a shared ruler that shows how far it was fooled. - **[A reaction time test](/tools/reaction-time-test).** Each student catches a falling ruler on screen and reads their time where they caught it. ## Keep it ethical: deception and the debrief These are demonstrations of published results, not new research, and a few rules keep them that way. **Keep wrong answers private.** The split-room experiments and the polls show the room only its totals. The traps quiz is scored, and its leaderboard shows the names people join with, so ask the room to join with nicknames if nobody should see who fell for a trap. The point is what the room did, not who fell for it. **Tell the room about the split straight away.** The two-version experiments hide one thing, that there were two versions. Say so at the reveal, and say why. **Never stage the cruel ones.** Conformity and obedience are taught by prediction. Nobody is lied to about what the group said or told to hurt anyone. **Ask before you publish anything.** If you might write up what your students did, check your institution's rules on research with students before you collect anything you plan to use. ## Running psychology experiments live with Decka Everyone joins from their own phone with a code, with no app and no account. The split-room experiments run on [A/B Tests](/ab-testing), which split the phones at random and show the two versions side by side before the results. The reasoning traps run as a [live quiz](/activities/live-quiz), where each reveal shows the share of the room that answered correctly. The predictions run on [live polls](/activities/live-polls), which show each answer's count and percentage as the votes land. The deck on this page runs them in the same three groups, with the published result on the slide after each and the source in the speaker notes. To run the framing experiment on its own, press "Make this A/B test free" at the end of this page. Decka runs inside [PowerPoint](/powerpoint) as an add-in and inside [Google Slides](/google-slides) through a Chrome extension, so the experiments can sit inside a lecture you already give. ## Frequently asked questions ### What are some psychology experiments for students? The framing effect, the anchoring bias and the decoy effect each split a class in two. The Linda problem, the Wason card task and the letter question catch the whole room at once. Asch's line study and Milgram's obedience study work as predictions, compared with what the originals found. ### What is a good framing effect example for class? Tversky and Kahneman's disease problem. Described as lives saved, [72% of 152](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) students chose a certain program over a gamble. Described as deaths, [22% of 155](https://www.csc2.ncsu.edu/faculty/mpsingh/local/Social/f24/wrap/readings/Tversky+Kahneman-framing-1981.pdf) did, though the outcomes were identical. ### What is an example of anchoring bias? In Tversky and Kahneman's study, people estimated the share of African countries in the United Nations after spinning a wheel. Groups that started from [10 and 65](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf) gave median estimates of [25 and 45](https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf). ### What is the Linda problem? A test of the conjunction fallacy. People read a description of Linda, then judge whether she is more likely to be a bank teller, or a bank teller active in the feminist movement. In Tversky and Kahneman's study, [85%](https://pages.ucsd.edu/~mckenzie/TverskyKahneman1983PsychRev.pdf) chose the second, which cannot be the more probable. Decka's [conjunction fallacy lesson](/classroom/psychology/conjunction-fallacy) runs it with a class. ### What is the Wason selection task? A logic puzzle with four cards showing a letter or a number, and a rule to test. In Wason's experiments, most students picked the cards that could confirm the rule rather than the ones that could break it. Across four of them, [5 of 128](https://iphils.uj.edu.pl/~a.trybus/classes/logic/cs/15.Wason.Natural.And.Contrived.Reasoning.pdf) students chose correctly. ### Can you run the Milgram experiment in class? No, and you should not try. Run it as a prediction: ask the class what share of Milgram's volunteers went to the end, then show that [26 of 40](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf) did, against a predicted [1.2%](https://www.psy.miami.edu/_assets/pdf/rpo-articles/milgram-1963.pdf). ### What is the difference between the availability heuristic and the representativeness heuristic? Availability judges how common something is by how easily examples come to mind, which is why words starting with R feel commoner. Representativeness judges how likely something is by how well it fits a picture, which is why Linda the activist feels likelier than Linda the bank teller. ## Activities - [Live Polls](https://decka.dev/activities/live-polls.md): Ask anything, watch votes land live - [Live Quizzes](https://decka.dev/activities/live-quiz.md): Timed questions, live leaderboard - [A/B Tests](https://decka.dev/ab-testing.md): Split the room, compare responses [Psychology experiments, live](https://decka.dev/templates/psychology-experiments-live-0c876e8645c9.md)