What is the role of recognition in decision making? Ben Newell University College London & Centre for Economic Learning & Social Evolution Acknowledgements:

What is the role of recognition What is the role of recognition in decision making?in decision making?

Ben NewellUniversity College London

&

Centre for Economic Learning & Social Evolution

Acknowledgements: David Shanks, Nicola Weston, Tim Rakow Funding: ESRC, Leverhulme Trust

Role of recognition in a cue-Role of recognition in a cue-learning tasklearning task

Previous work examined empirical evidence for building blocks of fast & frugal heuristics (e.g., search, stopping, decision rules) in menu-based tasks

Natural extension – examine evidence for fundamental ‘building block’ – the use of recognition

What do we mean by recognition?What do we mean by recognition?

Distinction between the truly novel and the previously experienced

E.g. nonwords – “prache”, “elbonics”Repetition of nonwords makes them

recognisable How much ‘weight’ is placed on simple

recognition?

Status of Recognition Information

Proportion of trials on which

recognized company chosen

Proportion of trials on which

advice is purchased

Special RH = RL RH = RL < NR

Consistent with other cues

RH > RL NR = RH < RL

RH = Recognition High (recognition best predictor of company performance) RL = Recognition Low (recognition poorest predictor of company performance) NR = No Recognition (Free advisor informational equivalent of RH)

Choices in accord with recognitionChoices in accord with recognition

Predictions: “special” RH = RL = 1.0; “consistent” RH > RL

00.10.20.30.40.50.60.70.80.9

1

RH RL NR

Pro

po

rtio

n o

f fi

nal 64 t

rials

Advice PurchaseAdvice Purchase

Predictions: “special” RH = RL (=0) < NR; “consistent” NR = RH < RL

00.10.20.30.40.50.60.70.80.9

1

RH RL NR

Pro

po

rtio

n o

f fi

nal 64 t

rials

Compensatory use of cuesCompensatory use of cues

Evidence for compensatory cue use in all conditions, most in RL

00.10.20.30.40.50.60.70.80.9

1

RH RL NR

Pro

po

rtio

n o

f fi

nal 64 t

rials

ConclusionsConclusions

Recognition information not ascribed “special status” in cue learning task

Treated as ‘just another cue’ in the environment (cf.,PROBEX Juslin & Persson 2002)

What about inferences from memory – do these rely on a ‘different sort of recognition’?

Recognition, Availability, Recognition, Availability, Familiarity…….Familiarity…….

Powerful influences on inferences from memory

Availability Heuristic (Kahneman & Tversky)

“Overnight Fame” effect (Jacoby)

Recognition Heuristic (Goldstein & Gigerenzer)

No Recognition Recognition +

Availability

(ease of recall)

Recognition

Recognition HeuristicRecognition Heuristic

Adaptive, non-compensatory, “all or nothing” use of recognition – “no other information is searched for”

Cities task with football team informationWhen is such a rule applied? What are

the consequences…..?

“Paying for the name…….”

Paying for the name…..Paying for the name…..

Hoyer & Brown (1990)– 3 brands of peanut butter, – “Aware group”:1 known, 2 unknown brands– 5 trials, opportunity to sample after each choice

Support for use of brand recognition in choice of peanut butter (DVD’s, computers, cars…….??)

% of participants choose known brand

Explicit (sole) use of brand awareness

‘heuristic’

Trial 1 93.5% 60%

Trial 5 74.5% 17%

Paying for the name…..Paying for the name…..

Hoyer & Brown (1990) contd…. Comparison with “No Awareness” group Significantly more sampling of brands in No Awareness group AND Awareness/quality-difference manipulation showed:

Reliance on Brand Awareness heuristic led to decreased search and final choice of inferior alternative

% of participants chose high quality brand when in an ‘unknown brand’ jar

Brand Awareness 20%

No Awareness 59%

““A good name is better than A good name is better than riches” (?)riches” (?)

Borges et al. (1999) – can “ignorance” beat the stock market? 180 German lay-people recognition of German stocks 6 month return on DAX 30: Dec 1996 – Jun 1997

Result replicated in 6 out of 8 tests Conclusion – ignorance can beat the stock market or big firms

do well in strong bull (up) markets?

Market Index Rec > 90% Rec < 10%

+34% +47% +13%

““A good name is better than A good name is better than riches” (?)riches” (?)

Boyd (2001) – test in a down or ‘bear market’ 184 US students recognition of 111 companies randomly

selected from Standard & Poor’s 500 6 month return: June 2000 – December 2000

No evidence to support use of recognition heuristic in a ‘bear’ market

Borges et al result a ‘big firm’ effect?

Market Index Rec > 90% Rec < 10%

-4.54% -14.75% +16.27%

““A good name is better than A good name is better than riches” (?)riches” (?)

Rakow (2002) – further test in a strong market 53 UK students recognition of 30 companies in Italian Mib 30 8 week return: October – December 2002 Compared recognition portfolios, anti-recognition portfolios and

expert portfolio with market index

Only 7 out of 53 recognition portfolios outperformed market index How robust is recognition heuristic as an investment tool?

Market Index Expert Recognition Portfolio

Anti-Recognition Portfolio

+21.0% +21.0% +13.3% +26.2%

When will recognition be When will recognition be accurate…?accurate…?

It depends on the domain…..

f(n) = 2(n / N) (N - n / N -1) + (N – n / N) (N – n – 1 / N – 1) ½ + (n / N) (n – 1 / N – 1)

f(n) = proportion correct inferences

= recognition validity

=knowledge validity

N = reference class of objects

n = recognized objects

(fast and frugal?)

Deliberate or automatic?Deliberate or automatic?

Automatic ‘feeling of familiarity’ + deliberate application of heuristic?

“I recognise it so I’ll choose it” – deliberate selection and use of a heuristic from the toolbox? (cf., Kahneman & Frederick, 2002)

Recognition + “relevance check”. Enron?

ConclusionsConclusions

Cue learning: recognition treated like other cues in the environment

Consumer research:Exploitation of reliance on recognition

Stock market investment: ‘big firm’ rather than recognition effect – how generalisable?

Where to next?Where to next?

Further tests of use of recognition in memory-based inference

Cost/benefit effects on use of recognition

Discussion of adaptive/maladaptive use of recognition

Further specification of the domains in which the ‘recognition heuristic’ applies