Behold LEGO Reenactments of Famous Psychology Experiments, as Imagined by Artificial Intelligence

Cog­ni­tive sci­en­tist Tomer Ull­man, head of Harvard’s Com­pu­ta­tion, Cog­ni­tion, and Devel­op­ment lab, may have inad­ver­tent­ly blun­dered into an untapped vein of LEGO Icon inspi­ra­tion when his inter­est in AI led him to stage recre­ations of famous psych exper­i­ments.

If you think Vin­cent Van Gogh’s Star­ry Night LEGO play­set is a chal­lenge, imag­ine putting togeth­er the AI-gen­er­at­ed play­set inspired by Yale psy­chol­o­gist Stan­ley Milgram’s 1961 obe­di­ence stud­ies, above.

Par­tic­i­pants in these stud­ies were assigned to play one of two parts — teacher or learn­er. Part­ner pairs were seat­ed in sep­a­rate rooms, acces­si­ble to each oth­er by micro­phones. The teacher read the learn­er a list of matched words they’d expect­ed to remem­ber short­ly there­after. If the learn­er flubbed up, the teacher was to admin­is­ter an elec­tric shock via a series of labelled switch­es, upping it by 15-volts for each suc­ces­sive error. The micro­phones ensured that the teacher was privy to the learner’s increas­ing­ly dis­tressed reac­tions — screams, des­per­ate protes­ta­tion, and — at the high­est volt­age — radio silence.

Should a teacher hes­i­tate, they’d be remind­ed that the para­me­ters of the exper­i­ment, for which they were earn­ing $4.50, required them to con­tin­ue. They also received reas­sur­ance that the painful shocks caused no per­ma­nent tis­sue dam­age.

Here’s the thing:

The teach­ers were inno­cent as to the experiment’s true nature. They thought the study’s focus was punishment’s effect on learn­ing abil­i­ty, but in fact, Mil­gram was study­ing the lim­its of obe­di­ence to author­i­ty.

The learn­ers were all in on the ruse. They received no shocks. Their respons­es were all feigned.

If our eyes don’t deceive us, the Mil­gram exper­i­ment that the AI imag­ines is even more extreme than the orig­i­nal. It appears all par­tic­i­pants, includ­ing those wait­ing for their turn, are in the same room.

As some­one com­ment­ed on Bluesky, the new social media plat­form on which Ull­man shared his hypo­thet­i­cal play­sets, “the sub­tle details the AI has got wrong here are the stuff of night­mares.”

AI’s take on the Stan­ford prison exper­i­ment seems more benign than the con­tro­ver­sial 1971 exper­i­ment that recruit­ed 24 stu­dent par­tic­i­pants for a filmed study of prison life to be staged in Stan­ford University’s psy­chol­o­gy depart­men­t’s base­ment, ran­dom­ly divid­ing them into pris­on­ers and guards.

AI’s faith­ful recre­ation of the LEGO fig­urines’ phys­i­cal lim­i­ta­tions can’t real­ly cap­ture the faux guards’ bru­tal­i­ty — mak­ing their pris­on­ers clean out toi­lets with their bare hands, strip­ping them naked, and depriv­ing them of food and beds. Their pow­er abus­es were so wan­ton, and the pris­on­ers’ dis­tress so extreme, that the planned dura­tion of two weeks was scrapped six days in.

It’s worth not­ing that all the stu­dent par­tic­i­pants came to the study with clean bills of phys­i­cal and men­tal health, and no his­to­ries of crim­i­nal arrest.

Far less upset­ting are the cog­ni­tive sci­ence exper­i­ment play­sets depict­ing the delayed grat­i­fi­ca­tion of the Stan­ford Marsh­mal­low Test and the selec­tive atten­tion of the Invis­i­ble Goril­la Test (both right above).

Ull­man also steered AI toward LEGO trib­utes to B.F. Skinner’s oper­ant con­di­tion­ing cham­ber and Mar­tin Seligman’s learned help­less­ness research (below).

No word on whether he has plans to con­tin­ue exper­i­ment­ing with AI-engi­neered LEGO play­set pro­pos­als fea­tur­ing his­toric exper­i­ments of psy­chol­o­gy and cog­ni­tive sci­ence.

Fol­low on Bluesky if you’re curi­ous. You’ll need to reg­is­ter for a free account and apply for an invite code, if you haven’t already… wait, are we set­ting our­selves up to be unwit­ting par­tic­i­pants in anoth­er psych exper­i­ment?

Hmmm…

Via Kot­tke

Relat­ed Con­tent 

The Psy­chol­o­gy Exper­i­ment That Shocked the World: Milgram’s Obe­di­ence Study (1961)

The Lit­tle Albert Exper­i­ment: The Per­verse 1920 Study That Made a Baby Afraid of San­ta Claus & Bun­nies

– Ayun Hal­l­i­day is the Chief Pri­ma­tol­o­gist of the East Vil­lage Inky zine and author, most recent­ly, of Cre­ative, Not Famous: The Small Pota­to Man­i­festo and Cre­ative, Not Famous Activ­i­ty Book. Fol­low her @AyunHalliday.


by | Permalink | Comments (0) |

Sup­port Open Cul­ture

We’re hop­ing to rely on our loy­al read­ers rather than errat­ic ads. To sup­port Open Cul­ture’s edu­ca­tion­al mis­sion, please con­sid­er mak­ing a dona­tion. We accept Pay­Pal, Ven­mo (@openculture), Patre­on and Cryp­to! Please find all options here. We thank you!


Leave a Reply

Quantcast
Open Culture was founded by Dan Colman.