Consequences of misspecified mental models: Contrasting effects and the role of cognitive fit

• Author: Nicolaj Siggelkow,Anoop Menon,Dirk Martignoni
• DOI: http://doi.org/10.1002/smj.2479
• Published date: 01 December 2016
• Date: 01 December 2016

Strategic Management Journal

Strat. Mgmt. J.,37: 2545–2568 (2016)

Published online EarlyView 15 February 2016 in WileyOnline Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2479

Received 25 May 2010;Final revision received25 October 2015

CONSEQUENCES OF MISSPECIFIED MENTAL

MODELS: CONTRASTING EFFECTS AND THE ROLE

OF COGNITIVE FIT

DIRK MARTIGNONI,1ANOOP MENON,2and NICOLAJ SIGGELKOW2*

1Institute of Management, Università della Svizzera italiana, Lugano, Switzerland

2Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A.

Research summary:Mental models, reecting interdependencies among managerial choice

variables, are not always correctly specied. Mental models can be underspecied, missing

interdependencies, or overspecied,containing nonexistent interdependencies. Using a simulation

model, we nd that under- and overspecication have opposite effects on exploration,and thereby,

performance. The effects are also opposite, depending on whether a manager controls all choice

variables. The mechanism underlying our results is a feedback loop: misspecied mental models

inuence managerial learning about the effectiveness of choices; this learning guides how the

environment is explored, which in turn, affects which information will be generated for future

learning. We explore implications of these results for strategic management and introduce the

notion of “cognitive t” between the mental model of the decision-maker and the strategic

environment.

Managerial summary:Managers often rely on mental models to guide their decision-making.

These mental models, however, are often misspecied, that is, more or less complex than the

situation managers are facing. Using a simulation model, we study the consequences of such

misspecied mental models. We nd that the performance implications of misspecied mental

models crucially depend on whether the manager controls all choice variables. We identify

situations in which simpler mental models are better than overly complex ones, and vice versa.

Copyright © 2015 John Wiley & Sons, Ltd.

INTRODUCTION

Managerial decision-making is often guided by

mental models: models that reect the expected

consequences of chosen activities, and the assumed

interdependence relationships either among these

activities, or between these activities and variables

not under the control of the decision-maker. For

instance, activities along the value chain might

interact with each other; likewise, some activities

Keywords: mental models; cognitive t; simulation; inter-

dependencies; learning

*Correspondence to: Nicolaj Siggelkow, 2000 SHDH, Wharton

School, Philadelphia, PA 19104. E-mail: siggelkow@wharton.

upenn.edu

Copyright © 2015 John Wiley & Sons, Ltd.

might be particularly valuable given certain envi-

ronmental conditions. Managerial mental models,

however, do not always accurately capture the true

interdependence structure. In some cases, manage-

rial mental models are cognitive simplications

of complex realities (Bettis and Prahalad, 1995;

Eisenhardt and Sull, 2001; Schwenk, 1984), that

is, mental models are underspecied in that they

ignore some existing interdependencies; they con-

tain causal “blind spots” (Zajac and Bazerman,

1991). In other cases, for example, through super-

stitious learning (Levinthal and March, 1981), man-

agers’ mental models may be overspecied, that

is, containing relationships that do not actually

exist.

2546 D. Martignoni, A. Menon, and N. Siggelkow

It is not surprising that managers’ mental models

are misspecied: Even for researchers who can rely

on large datasets, identifying and statistically vali-

dating interdependence relationships is a very dif-

cult undertaking (Athey and Stern, 1998), and has

led to many conicting results, for instance, in the

contingency literature (Donaldson, 2001). If men-

tal models are not accurate, the question of what

effects are caused by different kinds of misspeci-

cations in mental models naturally arises. What are

the consequences of having a mental model that is

overspecied, that is, assuming interdependencies

that do not exist? What are the consequences of

having a mental model that is underspecied and

blind to interdependencies that actually do exist?

Under which conditions is over- and underspecica-

tion particularly costly? For instance, does it make

a difference whether the interdependencies that are

misperceived are externalinterdependency relation-

ships, in which the value of an organizational choice

is affected by a variable that is not under the con-

trol of the manager, or are internal interdependency

relationships, in which the value of an organiza-

tional choice is affected by another variable that is

under the control of the manager? To start address-

ing these questions, we build a simulation model

that allows us to systematically analyze the perfor-

mance implications of having different misspeci-

cations within mental models.

To be more precise, we dene mental mod-

els as comprising of two components, the inter-

dependence representation that captures an indi-

vidual’s beliefs about the interdependencies among

the elements in the mental model, and the per-

formance representation that captures the assumed

mapping from actions to performance outcomes. In

the remainder of the article, for the sake of conve-

nience, we use the phrase under-/overspecied men-

tal models instead of under-/overspecied interde-

pendence representations within the mental models.

A short example will clarify the types of sit-

uations we analyze. Consider a manager who

contemplates how to best promote a new product

introduction. The manager may have a number of

alternatives: run a TV commercial, place an ad in

a magazine, or distribute yers in the streets. For

a new product, the effectiveness of these different

promotion choices may not be known to the

manager at the outset. In more abstract terms, the

performance landscape that the manager is facing is

not known to him or her. To make a decision about

which alternative to choose, the manager could

decide to run some experiments. For instance, the

manager could have yers distributed and observe

subsequent product sales. Given idiosyncracies on

any given day, the signal the manager receives about

the effectiveness of distributing yers will be noisy.

Over time, however, after running this experiment

repeatedly,the manager would be able to reduce this

noise. Likewise, the manager might run several TV

commercials and obtain an estimate of their effec-

tiveness. In this way,the manager will slowly create

aperformance representation of the actual perfor-

mance landscape he or she is facing, that is, the man-

ager develops a subjective performance landscape

reecting the manager’s beliefs concerning the

performance implications of the possible actions.

Now,consider an external contingency factor, the

weather, for example. The manager might expect

that certain interdependence relationships exist, for

instance, that the weather has an impact on the

effectiveness of distributing yers in the street

(sunny weather may improve mood, and hence, the

likelihood of people looking at yers). At the same

time, the manager may believe that the weather does

not have any impact on the effectiveness of TV

ads. Thus, the manager possesses an interdepen-

dence representation, a cognitive representation of

the assumed interaction structure between the var-

ious marketing alternatives and the (external) con-

tingency factor “weather” that is not under his or her

control.

Whether the weather actually has an impact on

the effectiveness of TV ads or yers is a different

issue. We could imagine that in reality the weather

has no systematic effect on any marketing alterna-

tive. In this case, the manager’s interdependence

representation (and mental model), which contains

an interdependence between weather and distribut-

ing yers, is overspecied— it includes an inter-

dependence relationship that does not really exist.

We could, however, also imagine that in reality the

weather has an impact on all promotion activities. In

this case, the manager’s mental model, which does

not contain the interdependence between weather

and TV ads, would be underspecied.

For the set-up with internal interdependencies,

imagine a slightly different situation in which

the manager has control over two organizational

choices: for instance, one decision among various

marketing alternatives and a second decision of

how to set the stafng level in the rm’s call center.

Again, the manager may have some beliefs about

whether these choices interact or not. For instance,

Copyright © 2015 John Wiley & Sons, Ltd. Strat. Mgmt. J.,37: 2545–2568 (2016)

DOI: 10.1002/smj