35
Evolved Decision Making Peter M. Todd Cognitive Science, Informatics, and Psychology, Indiana University & Center for Adaptive Behavior and Cognition (ABC), MPI, Berlin with Peter DeScioli and Rob Kurzban Max Planck Institute for Human Development Berlin

Evolved Decision Making

Embed Size (px)

DESCRIPTION

Evolved Decision Making. Peter M. Todd Cognitive Science, Informatics, and Psychology, Indiana University & Center for Adaptive Behavior and Cognition (ABC), MPI, Berlin with Peter DeScioli and Rob Kurzban. Max Planck Institute for Human Development Berlin. What should I eat?. - PowerPoint PPT Presentation

Citation preview

Page 1: Evolved Decision Making

Evolved Decision Making

Peter M. ToddCognitive Science, Informatics, and

Psychology, Indiana University& Center for Adaptive Behavior and

Cognition (ABC), MPI, Berlin

withPeter DeScioli and Rob Kurzban

Max Planck Institute for Human Development Berlin

Page 2: Evolved Decision Making

What should I eat?

Page 3: Evolved Decision Making

How can I decide?

152 types of street food, ~500 pages, 1200+ vendor reviews…

Page 4: Evolved Decision Making

Visions of Cognition

Superhuman

unbounded

rationality

BoundedRationalit

y

The Adaptive Toolbox

Page 5: Evolved Decision Making

A collection of decision mechanisms, including simple heuristics and rules of thumb...

Some evolved/pre-wired, some learned...

Applicable to specific decision tasks and in particular domains...

That rely on and exploit the structure of information in the environment...

To yield ecological rationality.

The adaptive toolbox

Page 6: Evolved Decision Making

For more info, see our book....

Also, see our new book coming soon:

Ecological Rationality:

Intelligence in the World

Page 7: Evolved Decision Making

Two types of heuristics

Choose between options by searching for as little information as possible

Search for the options themselves, stopping with as little search as possible

Page 8: Evolved Decision Making

Which US city has more inhabitants,San Diego or San Antonio?

Page 9: Evolved Decision Making

Which US city has more inhabitants,San Diego or San Antonio?

Americans:62% correct

Germans:?

correct

Germans:100%correct

(but not any more…)

Page 10: Evolved Decision Making

Definition: To decide between two objects on some dimension, if you recognize only one, pick it

This only requires search for recognition knowledgeEcological Rationality: This heuristic works well when recognition knowledge is correlated with the decision dimension

The Recognition Heuristic

Page 11: Evolved Decision Making

Animals use recognition heuristics

Evolutionary roots of recognition use:

Rats use recognition to decide whether to try a new food: try it if they recognize it from another rat

Bulls use anti-recognition in mate choice:

Fish, birds, and humans use mate copying

Page 12: Evolved Decision Making

“Paying for the name…….”

Page 13: Evolved Decision Making

Which city is bigger?

Dresden LeipzigCuesRecognized? yes yesNational capital no noSoccer team no yesIntercity train yes yesState capital no yesUniversity yes yes

How to decide? Weight/add or tally all available cues, or use just one?

Page 14: Evolved Decision Making

Search: How to find things

How do we choose among options that are not all present to us at once?

Search is required whenever resources are distributed in space or time, e.g.:

• food• mates, friends• house/apartment• jobs

Page 15: Evolved Decision Making

The problem of finding things

When searching, another better alternative could always be found, so the real problem is:

when to stop search—exploit the current item, and when to keep looking—explore further?

This explore/exploit tradeoff in an ongoing dynamic process—also in business

Evolution has built mechanisms to explore quickly and exploit as much as possible

Page 16: Evolved Decision Making

Features of employee search

No going back: once an alternative is passed, there’s little chance of returning to it

No looking forward: upcoming range of possible alternatives is largely unknown

Mutual search: You must choose an employee, and convince him/her to join your company

How to decide when to stop and “exploit” the current option, or explore for someone else?

Search using an aspiration level, or satisfice (Herbert Simon, 1955)

Page 17: Evolved Decision Making

Heuristics for mutual search

Use an aspiration level dependent on one’s own market value to speed up mutual search

But how to determine one’s own value on the job market in a quick and simple way?

Answer: learn one’s own value during an initial “testing” period based on unexpected feedback, and use this as aspiration level

Page 18: Evolved Decision Making

Testing search rules empirically

What rules are people likely to use in such mutual sequential searches? Those we have evolved to use in related settings, e.g. mating

How can we observe ongoing mate choices in individuals, without having to follow people for years?

We need sped-up mate choice in a microcosm...

Page 19: Evolved Decision Making

Speed dating...

Page 20: Evolved Decision Making

How does speed-dating work?

• ~20 men and ~20 women gather in one room (after paying $30)

• Women sit at tables, men move in circle• Each woman talks with each man for 5 min.• Both mark a card indicating whether they

want to meet the other ever again• Men shift to the next woman and repeat• (now happening worldwide, with many

variations)

Page 21: Evolved Decision Making

What happens next

• Men’s/women’s “offers” are compared

• Every mutual offer gets notified by email, with other’s contact info

• After that, it’s up to the pairs to decide what to do

To study mutual search heuristics, we ran our own speed-dating sessions….

Page 22: Evolved Decision Making

Separate “dating” booths

Page 23: Evolved Decision Making

One of the booths (with camera, mics)

Page 24: Evolved Decision Making

A “date” in progress

Page 25: Evolved Decision Making

How people searchedSpeed-daters appeared to follow an aspiration-

adjustment heuristic in deciding to make offers to people they expected offers from

Similar heuristics may be used in organizational search problems like hiring or finding corporate partners

Page 26: Evolved Decision Making

Selecting the right tool

Problem associated with having lots of tools in the toolbox: When to pick each to use?

However it works, the tool-selection mechanism must be simple, fast, & frugal too

Mind has evolved to detect cues for different types of problems and choose proper tools

Cues in workplace can lead people to think they’re in a friendship/cooperative setting or anonymous competitive setting and choose decision tools accordingly

Page 27: Evolved Decision Making

Conclusions• Organizations as goal-seekers must make

good decisions within constraints

• Good choices can be made with little info processed by appropriate tools from toolbox

• Tools include heuristics for choice, allocation, and search for resources

• Selecting tools must be done on basis of cues from environment, again simply

• Can we set up work environments so they elicit decision tools we want on the job?

Page 28: Evolved Decision Making
Page 29: Evolved Decision Making

How to be rationalGiven our mental constraints (Herbert Simon):

– Human memory, cognitive capacity, and time are limited, i.e. bounded

– Moreover, most real-world problems cannot be solved by optimization

The only plausible/possible approach is:– to make decisions within our bounds by using

approximate methods: heuristics, shortcuts, “rules of thumb”, that make “quick and dirty” decisions

– We can’t use all available information, so we have to search for (limited) information to use

Page 30: Evolved Decision Making

A well-studied search example: the Secretary Problem

A company faces this challenge:• 100 job applicants with unknown distribution of

typing speeds will be seen• Applicants are interviewed in sequence and

announce their typing speed• Search can be stopped at any time by hiring the

current candidate, but no returning to earlier applicants

How can the company maximize its chances of hiring the best secretary?

Page 31: Evolved Decision Making

Solving the Secretary Problem

Goal: Maximize chance of finding best option

Approach: Set aspiration level by sampling a number of options that balances information gathered against risk of missed opportunity

Solution: Sample N/e (= .368*N), or 37% Rule (set aspiration after seeing first 37% of apps)

Other approaches: maximize mean return with a much shorter initial search period (~10%)

Page 32: Evolved Decision Making

Implications

• People will aim to stop their search for options quickly

• People will use feedback about what they can “afford” to adjust their aspirations (so we can shape their decisions by ordering options)

• People will use social info from others to adjust their own preferences and shorten their search

• In these ways, (too) many options need not scare off decision makers

Page 33: Evolved Decision Making

Examples in Games

• Double auction: buyers/sellers set prices for goods—behavior is rational and prices converge quickly– Cues indicate that cost/benefit should be used

• Public goods: players can keep money or contribute to “public account”, get more back– Cues to social interaction/group identity, e.g.

glances, cheap talk, punishment, lead to more giving to public account

• Dictator game: eye spots induce more giving

Page 34: Evolved Decision Making

Examples in social relations

Organizations contain many types of social relations: friends, mates, trade partners, leader-follower, boss-worker, etc.

Crucial to know what relationship you’re in

Cues can signal relationship type

E.g. friendship (vs. exchange partner) signified by friend ranking you above others (DeScioli & Kurzban 2009)

Page 35: Evolved Decision Making

ConclusionsThe ecological rationality approach to studying simple

decision heuristics emphasizes the match between:• heuristics specified in terms of info search,

stopping, and decision building blocks• how structured information in the decision

environment can be exploited by heuristics

This approach (with multiple methodologies) helps us predict how people make decisions, e.g. about what to eat, and influence choices by constructing appropriately-structured environments