LAUREIRO Etal 2010 Exploration Neuro JNPE

8/6/2019 LAUREIRO Etal 2010 Exploration Neuro JNPE

1/21

The Neuroscientific Foundations of the ExplorationExploitationDilemma

Daniella Laureiro-MartnezBocconi University and Universidad de los Andes

Stefano Brusoni and Maurizio ZolloBocconi University

What are the origins of the ability to continuously explore novel domains of activitywhile at the same time exploiting the current knowledge base with increasing efficacy?The conflicting objectives of exploration and exploitation compete for scarce resources,among which managerial attention is possibly the most critical. This paper integratesrecent findings on the neuromodulation of attention to provide a foundational step inunderstanding how the mind of the manager handles the explorationexploitationdilemma. Also, this paper proposes several possible ways to combine research in

neuroscience, psychology, and management to advance our knowledge of the micro-foundations of managerial decision-making.

Keywords: microfoundations, exploration, exploitation, attention modulation

Adaptive firm behavior in a diverse and rap-idly changing environment requires a trade-offbetween exploiting known sources of rewardand exploring the environment for more valu-able or stable opportunities. This trade-off isknown as the explorationexploitation di-

lemma (March, 1991) and is present at differentlevels of analysis and different time scales ofdecision-making. There is no general optimalpolicy for how to manage the trade-off betweenexploration and exploitation in nonstationaryenvironments (Cohen, McClure, & Yu, 2007).

From a managerial point of view, theexplorationexploitation dilemma, which iskey to many of the challenges faced by organi-zations, refers to the difficulty faced by organi-

zations and their members in trying to find abalance among competing activities in the con-text of scarce resourcesthe need to be effi-cient to get the most from a current situation,while at the same time exploring possibilitiesfor future improvements. As March (1991) puts

it:Adaptive systems that engage in exploration

to the exclusion of exploitation are likely to findthat they suffer the costs of experimentationwithout gaining many of its benefits. They ex-hibit too many undeveloped new ideas and toolittle distinctive competence. Conversely, sys-tems that engage in exploitation to the exclusionof exploration are likely to find themselvestrapped in suboptimal stable equilibria. As a re-sult, maintaining an appropriate balance between

exploration and exploitation is a primary factor insystem survival and prosperity (March, 1991).Levinthal and March (1993) similarly argue

that an organization that engages exclusivelyin exploration will ordinarily suffer from thefact that it never gains the returns of its knowl-edge and that an organization that engagesexclusively in exploitation will ordinarily sufferfrom obsolescence (p. 105). A narrow searchcan lead to increasingly rigid cognitive mapsand highly specialized competencies that maybecome core rigidities (Leonard-Barton, 1995).The so called ambidexterity hypothesis(Tushman & OReilly, 1996) states that thehigher the organizational ability to balance ex-

Daniella Laureiro-Martnez, Department of Management,Bocconi University, and School of Management, Univer-sidad de los Andes; Stefano Brusoni, KITeS-CESPRI, De-partment of Economics, Bocconi University; and MaurizioZollo, CROMA, Department of Management, Bocconi Uni-versity.

We thank Nicola Canessa, Stefano Cappa MD, LuisFelipe Orozco-Cabal MD, David Papo and the participantsin the seminars at the IRI Summer School, the BocconiManagement Seminars, and the JNPE and SMS Confer-ences for their comments and suggestions. Any mistakes inthis article are the sole responsibility of the authors.

Correspondence concerning this article should be ad-dressed to Daniella Laureiro-Martnez, Via Roentgen, 1,Office 5 B1-12, 20136 Milan, Italy. E-mail: [email protected]

Journal of Neuroscience, Psychology, and Economics 2010 American Psychological Association2010, Vol. 3, No. 2, 95115 1937-321X/10/$12.00 DOI: 10.1037/a0018495

95


2/21

ploration and exploitation, the higher the orga-nizational performance.

Since Marchs (1991) seminal article, themanagement literature has used the terms ex-

ploration and exploitation in studies of orga-nizational adaptation, organizational learning,competitive advantage, technological innova-tion, organization design, and organizationalsurvival. However, an examination of the lit-erature indicates that the answers containedthere to the central questions on this subjectremain incomplete, at times contradictory, andat best ambiguous (Gupta, Smith, & Shalley,2006). The review in Section 2 shows that whilethe explorationexploitation literature is exten-

sive and growing, there are gaps that we believederive from a lack of understanding of the di-lemma at the micro level, that is, the individualdecision-makers point of view. There is anintriguing opportunity to better understand thispoint of view. The recent development ofknowledge in the cognitive neurosciencesopens, in fact, exciting possibilities to buildintegrative approaches to understand organiza-tional dilemmas from a micro perspective. Thatis the overarching objective in this paper: toexplore an organizational conundrum from anindividual level of analysis, showing how inter-personal variation in the decision-makers neu-rological disposition affects the decisional out-comes and, potentially, the performance of agiven task connected to that decision.

To do so, the paper is organized as follows:Section 2 briefly presents the main gaps in theorganizational literature on explorationexploita-tion, focusing on the one that this paper aims tocontribute to; Section 3 addresses the micro-foundationsthat is, the individual originsof

this dilemma, whereas Section 4 contributes bybridging with neuroscience and discussing cer-tain findings in that domain that may help toclarify the roots of the managerial dilemma andsuggest ways to cope with it. We then providean illustration of how the theory we developedmight explain the behavior and the outcomesconnected to one of the most famous innovatorsof modern times: Thomas Alva Edison. Sec-tion 6 then presents the main challenges in-volved in the development of an empirical

agenda to validate these ideas, offering a full setof suggestions on how to tackle them, and Sec-tion 7 concludes with several suggestions for

the development of this line of work in futureresearch.

Gaps in the Management Literature on the

ExplorationExploitation Dilemma

A review of the literature on theexplorationexploitation managerial dilemmauncovers several limitations and gaps, which webriefly cover with the aim to focus on andpotentially contribute toward resolving one par-ticular gap: variation among individuals in thetendency to respond in an exploitative or ex-plorative way to a given stimulus, and on theability to shift their responses according tochanges in the environmental conditions, havenot been explored.

A central gap is that most of the extant re-search focuses on the structural antecedents toand the effects of involvement in explorationand exploitation on firm performance (Raisch &Birkinshaw, 2008). Few studies delve moredeeply into how these two activities occur si-multaneously. Of course, part of the how questionhas to do with the actual capacities and behaviorsofindividual members of the organization, ratherthan with organizational arrangements and collec-

tive processes. Another key gap is the lack ofclarity about the levels of analysis in research onexplorationexploitation. Recent theoreticalcontributions ( Felin & Foss, 2005; Felin, &Hesterly, 2007; Rothaermel & Hess, 2007)identify two main problems with the single-level research approach. First, focusing on onlyone level of analysis implicitly assumes thatmost of the heterogeneity is located at that levelwhile other levels are more or less homoge-neous. Second, this focus implies that this levelis independent of interactions with other lower-or higher-order levels of analysis.

Another main concern with the current liter-ature is the lack of agreement about key ele-ments regarding the definitions of explorationand exploitation. It is not clear from the litera-ture whether exploration and exploitationshould be viewed as two ends of a continuum,or as two different and orthogonal aspects oforganizational behavior. The central ambiguityin the definitions is whether exploration andexploitation differ in the type of learning or by

the presence/absence of learning (Gupta et al.,2006). Table 1 summarizes some of the defini-tions that appear in key articles on the subject.

96 LAUREIRO-MARTINEZ, BRUSONI, AND ZOLLO


3/21

In order to avoid confusion, in this paper weadopt a definition very close to the one providedin neuroscience (discussed in Section 4) thatadmits learning in both exploration and exploi-tation. We define exploration as the behaviorthat includes search for alternatives and disen-gagement from the current task. The simplestform of exploration is random search but othermore structured types of search, such as the useof heuristics or explicit algorithms, are alsoincluded. As a consequence of this behavior,experimentation, flexibility, discovery, and in-novation are shown. We define exploitation asthe behavior that helps optimize task perfor-mance. When this behavior is present, there is ahigh engagement with the current task. As aconsequence of this behavior, selection, refine-ment, choice, production, and a concern with

efficiency are shown.The fourth major gap in the received litera-

ture on the explorationexploitation dilemma,

and the one we focus on, concerns the role ofthe characteristics of individual traits as mech-anisms underpinning the development of orga-nizational capabilities related to the balancedmanagement of exploration and exploitation ac-tivities. The roles of individual and group char-acteristics were viewed as an important andnecessary focus of scholarly attention since theinception of the behavioral theory of the firm(Cyert & March, 1963; March & Simon, 1958;Simon, 1985). March (1991), for example, de-scribes the cognitive limits that individuals en-counter when trying to conduct explorative andexploitative processes simultaneously. AndTushman and OReilly (1996) argue, the abilityto exploreexploit at the organizational level isfacilitated by the top-management teams inter-nal processes.

It is only recently, however, that scholarshave started to empirically investigate teamcharacteristics that enable organizations to man-

Table 1Definitions of Exploration and Exploitation

Quotes from: Exploration Exploitation

March, 1991 Exploration includes elements captured by

terms such as search, variation, risk-taking,experimentation, play, flexibility, discovery,and innovation.

Exploitation includes such things as refinement,

choice, production, efficiency, selection,implementation, and execution. The essenceof exploitation is the refinement andextension of existing competences,technologies, and paradigms. Its returns arepositive, proximate, and predictable.

The essence of exploration is experimentationwith new alternatives. Its returns areuncertain, distant, and often negative.

March, 2006 Pursuit of what might come to be known. The refinement and implementation of what isknown.

Holmqvist, 2004 Exploration is concerned with creating varietyin experience, and thrives onexperimentation and free association.

Exploitation is about creating reliability inexperience, and thrives on productivity andrefinement.

Variety in experience through search,discovery, novelty, innovation, and

experimentation.

Creates reliability in experience throughrefinement, routinization, production, and

implementation of knowledge.Levinthal and Rerup,2006

Experimenting with a novel action impliesforgoing the use of existing, establishedpractices. In this sense, mindfulnesscorresponds to exploratory behavior . . .

. . . and less-mindful behavior is akin toexploitative behavior.

Zollo and Winter,2002

Exploration activities are primarily carried outthrough cognitive efforts aimed atgenerating the necessary range of newintuitions and ideas (variation), as well asselecting the most appropriate ones throughevaluation and legitimization processes.

By contrast, exploitation activities rely more onbehavioral mechanisms encompassing thereplication of the new approaches in diversecontexts and their absorption into theexisting sets of routines for the execution ofthat particular task.

Smith and Tushman,2005

Exploration is rooted in variance-increasingactivities, learning by doing, and trial anderror; exploration creates futures that may

be quite different than the organizationspast.

Exploitation is rooted in variance-decreasingactivities and disciplined problem-solvingexploitation builds on an organizations past.

Levinthal andMarch, 1993

the pursuit of new knowledge of things thatmight come to be known.

the use and development of things alreadyknown.

97NEUROFOUNDATIONS EXPLORATIONEXPLOITATION DILEMMA


4/21

age both exploration and exploitation (Beck-man, 2006; Lubatkin, Simsek, Ling, & Veiga,2006; McKenzie, Woolf, van Winkelen, &Morgan, 2009; Mom, Van Den Bosch, & Vol-

berda, 2009; Perretti & Negro, 2006; Smith &Tushman, 2005). Gibson and Birkinshaw (2004,p. 223), for example, note the important roleplayed by senior executives in making an orga-nization context effective and developing ambi-dexterity (p. 223). Similarly, Smith and Tush-man (2005) explore the integrative mechanismsby which leadership teams might successfullymanage the contradictions that arise from struc-tural separation in ambidextrous organizations,and Volberda, Baden-Fuller, and Van DenBosch (2001) note that top management ex-plicitly manages the balance of exploration andexploitation by bringing in new competencies tosome units while utilizing well-developed com-petencies in others (p. 165). And at the grouplevel of analysis, Beckman (2006) finds evi-dence that the composition of the foundingteam, and members prior company affiliationsin particular, is an important antecedent tofirms exploitative and explorative behaviors.

Unfortunately, the focus on team characteris-tics as the antecedents to the development of

organizational capabilities in ambidexterity isnot matched by studies on the role of managersindividual characteristics or on the ability tomake balanced explorationexploitation deci-sions. OReilly and Tushman (2004), for in-stance, emphasize the role of ambidextrousmanagers with the ability to understand and besensitive to the needs of very different kinds ofbusinesses (p. 81). Despite a seeming consen-sus on the importance of the individual, the stateof the art in scholarly work on the individuallevel of analysis shows a concerning paucity,with the notable exception of Mom, Van DenBosch, and Volberda (2007). In our view, this isan important concern, as many current holis-tic explanations might capture at least some ofwhat really are effects of variation at the indi-vidual level of analysis (Felin & Foss, 2005;Felin & Hesterly, 2007). We explore this furtherin Section 3.

The ExplorationExploitation Dilemma in

the Managers Mind

Most studies of explorationexploitation, start-ing with Marchs (1991) seminal work, focus on

levels of analysis above the individual. March(1991) analyzes the explorationexploitationtrade-off in the social context of organizations,focusing on two distinctive features: mutual

learning in the organization and the individualsinvolved, and the competition for primacyamong organizations. Implicit in his model isthe assumption that the balance between explo-ration and exploitation is based on a turnoverprocess among a mix of individuals with differ-ent cognitive characteristics (some more in-clined to exploration, others more inclined toexploitation), who achieve a trade-off for thewhole organization.

The mechanism based on the turnover ofcognitively specialized (and inflexible) man-agers to achieve a balance between those pre-disposed to exploitative behavior and those pre-disposed to explorative behavior, however, isclearly not the only one at the disposal of theorganization (the alternative, of course, is togather a group of managers who can think andact in both modes with relative ease), and prob-ably not even the optimal one. First of all,having a group of cognitively specialized deci-sion-makers will simply turn the decision into apolitical battle between the two factions, with

results driven by the relative size and politicalweight of the two factions, rather than to ratio-nal choice. Second, even when the decision canbe efficiently allocated to exploration-minded or to exploitation-minded manag-ers, this decision itself requires a decision-maker (e.g., the CEO) with a significantamount of cognitive flexibility to recognizethe advantages and disadvantages of the twoalternative allocations of the decision-makingresponsibility. Essentially, the problem (andthe solution, via cognitive flexibility) wouldstill be there, but would be upgraded to thecognitive profile of the person at the top ofthe organization, who assigns the problem tothe cognitively specialized groups.

The alternative scenario of having cogni-tively nonspecialized (and flexible) managersmight thus be superior to the one describedabove, since it would reduce the likelihood ofpolitical fights and suboptimal decisions andwould not require the presence of a higherorder decision-maker to assign the responsi-

bility of the solution to cognitively special-ized groups of managers. As OReilly andTushman (2004) recognized, one of the most



5/21

important lessons is that ambidextrous orga-nizations need ambidextrous senior teams andmanagers (p. 81).

We thus propose that to achieve a better

understanding of how both exploration and ex-ploitation can be conducted, we need an in-depth examination of the microfoundations oforganizational learning. The microthat is, in-dividuallevel of analysis may in fact accountfor an important amount of heterogeneity indecisional and performance outcomes, andshould be explicitly studied:

to fully explicate organizational anythingwhether identity, learning, knowledge or capa-bilitiesone must fundamentally begin withand understand the individuals that compose thewhole, specifically their underlying nature,choices, abilities, propensities, heterogeneity,purposes, expectations and motivations. Whileusing the term organizational may serve ashelpful shorthand for discussion purposes andfor reduced-form empirical analysis, truly ex-plaining the organization (e.g., existence, de-cline, capability or performance), or any collec-tive for that matter, requires starting with theindividual as the central actor (Felin & Foss,2005).

Mom et al. (2007) were, to the best of ourknowledge, the first to analyze the explorationexploitation dilemma at the individual level ofanalysis. They explored the influence of knowl-edge flows on the managers explorative or ex-ploitative activities, recognizing that one of themost promising avenues for future research ismeasuring exploration and exploitation at themanagerial level of analysis using objectivemeasures (p. 927).

Along these lines of discourse, Section 4 dis-cusses recent findings on the neuromodulationof attention, which, we argue, is at the core ofthe human ability to shift from one learningmode to another. We provide some suggestionson how to build on the concepts of situationaluncertainty, utility perception and attention fo-cus to further investigation of organizationalexploration and exploitation.

A Neuroscientific Approach to the

ExplorationExploitation Dilemma

In line with our aim to examine the microfoun-dations of the explorationexploitation dilemma,we searched for work on the cognitive processes

underlying the explorationexploitation trade-off.Recent work on the neuromodulation of attentionproposes that interactions between the orbitofron-tal cortex (OFC), the anterior cingulate cortex

(ACC), and the locus coeruleus (LC) (seeFigure 1) may modulate attention and thus bal-ance explorationexploitation (Aston-Jones &Cohen, 2005; Cohen, Aston-Jones, & Gilzenrat,2004; Usher, Cohen, Servan-Schreiber, Ra-

jkowski, & Aston-Jones, 1999).Traditionally, the locus coeruleusnorepineph-

rine (LC-NE) system was considered to be im-plicated solely in arousal. However, multiplerecent findings suggest that this system plays amore complex and specific role in the control ofbehavior by contributing to the optimization ofbehavioral performance (Sara, 2009).

Aston-Jones, Rajkowshi, Kubiak, and Alex-insky (1994) observe that the LC shifts betweentwo operating modes: the phasic and the tonic.In the former, LC cells exhibit phasic activationin response to the processing of task-relevantstimuli, but display only a moderate level oftonic discharge. This mode is consistently asso-ciated with enhanced attention focus, generatingexploitative behavior, defined as behaviorthat optimizes and achieves high levels of task

performance (Aston-Jones & Cohen, 2005). It isimportant to note here that the analysis is at theindividual level. Exploitative behavior trans-lates into a high level of engagement with thecurrent task. As a consequence, behaviors thatshow refinement in the selected actions and aconcern for efficiency are demonstrated.

In the tonic mode, LC cells do not respondphasically to task events, but exhibit higher

LC

OFC

Figure 1. LC, OFC and ACC location in the brain.



6/21

levels of ongoing tonic activity. Exploration isdefined as the behavior shown when there issearch for alternatives and disengagement withthe current task (Aston-Jones & Cohen, 2005).

This mode is associated with explorative behav-ior because it corresponds with poor perfor-mance on tasks that require focused attention,and with increased distractibility. The simplestform of exploration is random search, but ex-ploration also includes more structured types ofsearch, such as the use of heuristics or explicitalgorithms. This behavior demonstrates abilitiesfor experimentation, flexibility, discovery, andinnovation.

It should be noted that the definitions of ex-ploration and exploitation provided by neuro-scientists (Aston-Jones & Cohen, 2005) arecompatible with those in the management liter-ature (see Table 1) and with Marchs (1991)definition: The essence of exploitation is therefinement and extension of existing compe-tences, technologies, and paradigms. Its returnsare positive, proximate, and predictable. Theessence of exploration is experimentation withnew alternatives. Its returns are uncertain, dis-tant, and often negative. (p. 81).

The findings on the functioning of the LC and

its consequent type of behavior are based ontwo pieces of evidence. The first comes fromexperiments on monkeys, which transitionedfrom a phasic to a tonic mode and then reversedwhen the new target was identified. This tran-sition requires that LC has the relevant infor-mation to determine when to switch betweenphasic and tonic modes, an important aspectthat we address in the section on perception ofutility (Usher et al., 1999). The second piece ofevidence derives from studies of humans mea-suring pupil diametersa good proxy for LCactivityand functional MRI (fMRI) experi-ments (see Beversdorf, White, Chever, Hughes,& Bornstein, 2002; Daw, ODoherty, Dayan,Seymour, & Dolan, 2006; and Sterpenich et al.,2006; among others). Usher and colleagues(1999) also develop a biophysically plausiblemodel of LC functioning that accounts for tran-sitions between the phasic and tonic modes interms of a single physiological variable (cou-pling between LC cells) and explains the impactof these shifts on task performance. In brief, the

model suggests that the phasic mode favorsexploitation by releasing norepinephrine (NE)when a task-relevant event occurs, thereby fa-

cilitating the processing of that event, whilesustained release of NE in the tonic mode in-discriminately facilitates the processing of allevents irrespective of their relevance to the cur-

rent task, thereby favoring exploration.Viewed from the perspective of attention, the

LC phasic mode supports the current control state(exploitation), while the LC tonic mode provokesa withdrawal of control from the current task,favoring the sampling of other behavioral goals(exploration). These changes between phasic andtonic modes are the basis for an understanding ofthe explorationexploitation dilemma from an at-tention perspective. When the utility derived froma given behavior is low in comparison to expec-tations, flexibility to change the attention focus,and thus the behavior from exploitation to explo-ration, is needed to explore the environment andsample different behaviors until new sources ofreward are discovered. This is the role played bydifferent modes of activity in the LC-ACC/OFCsystem. Aston-Jones and Cohen (2005) observedthat the LC shifts between two distinct operatingmodes, and that these shifts change attention andthen behavior. The LC phasic mode supports thefocus of attention on the current control state (ex-ploitation), while the LC tonic mode provokes a

withdrawal of control from the current task, thusfavoring broader attention and the sampling ofother behaviors (exploration).1 The phasic LC re-sponses facilitate context-congruent behavioral re-sponses (exploitation) and the tonic mode of LCfacilitates sensitivity to different stimuli and theexecution of a broader class of behavioral re-sponses (exploration).

These neuroscientific findings on changes inattention scope contribute to our understandingof the management explorationexploitation

dilemma from an attention perspective. Theidea that broad attention is important in situa-tions that are dynamic, ill-structured, ambigu-ous, and unpredictable is acknowledged in themanagement literature (Levinthal & Rerup,2006; Weick & Sutcliffe, 2006). In the oppositescenarios, the economies of narrow attention aremore appropriate and, through their reliance onroutines, can be cost-efficient (Nelson & Win-

1 We should note here that, whereas we have described

the phasic and tonic modes as distinct, they likely representthe extremes of a continuum of function. For expositorypurposes it is more useful to distinguish them at the twoextremes.



7/21

ter, 1982). As a consequence, the higher theuncertainty in a situation, the higher the likeli-hood that broad attentionand thus explorativebehaviorswill lead to better performance.

However, these explanations still leave outthe important question of what information theneural system uses to determine whether itshould exploit (LC phasic mode) or explore (LCtonic mode). Studies have found that the braincomputes the perceived utility in a particularsituation, compares that with its expectationlevels, and drives shifts between LC phasic andtonic modes by influencing simple physiologi-cal parameters (Usher et al., 1999). When lowexpected utility is perceived, the broad mode ofattention (tonic mode of the LC functioning) isactivated. When high utility is perceived, thefocused mode (the phasic LC mode) is acti-vated. Studies of humans (and neuronal recordsin primates) show that the frontal cortex playsan important role in the evaluation of utility.Different areas of the brain are found to berelated to the assessment of rewards and costs.A large number of neuroimaging studies, in-volving diverse experiments, have examinedbrain responses to reward stimuli. They consis-tently identify a common set of neural structures

that are activated in response to these stimuli,mainly the OFC, the ventral striatum, and theventromedial prefrontal cortex. The OFC hasbeen implicated in hedonic experience across allsensory modalities (Rolls, 2000). Of specificinterest are areas in the striatum and the OFCthat are particularly responsive to rewards andwhich change, accumulate, or are learned overtime (Knutson, Fong, Bennett, Adams, & Hom-mer, 2003; Koepp et al., 1998; Murray,ODoherty, & Schoenbaum, 2007).

Attempts have also been made to identify thebrain areas activated by cost-related issues. Theanterior cingulate cortex (ACC) is traditionallyconsidered to be directly responsive to aversiveinteroceptive and somatosensory stimuli, andparticularly to pain (e.g., (Peyron, Laurent, &Garca-Larrea, 2000). Recent neurophysiologi-cal studies on monkeys and on humans haveconsistently demonstrated that ACC is stronglyresponsive to negatively valenced information,such as performance errors, negative feedback,monetary losses, and even social exclusion

(Eisenberger, Lieberman, & Williams, 2003),and also to task difficulty and decision-makingconflicts (Botvinick, Cohen, & Carter, 2004).

Therefore, there is much evidence to suggestthat the OFC is involved in reward evaluation,whereas the ACC is responsive to a variety ofnegatively valenced signals (Aston-Jones & Co-

hen, 2005; McClure, Laibson, Loewenstein, &Cohen, 2004; Montague, King-Casas, & Cohen,2006; Tom, Fox, Trepel, & Poldrack, 2007).These results point to the existence of a strongrelationship between situation uncertainty andutility perception. Thus, risk-averse individuals,all else being equal, will find higher utility inless uncertain situations and, as a consequence,will act under the phasic LC mode, showingmore exploitative behavior. In highly uncertainsituations, the opposite will occur. Individualswill find less utility in such situations and willshift to a tonic LC mode, acting in a moreexplorative way.

That perceiving a high utility reduces theattention to search and exploration, and that theperception of low utility promotes search andexplorative behavior, are fundamental tenets ofthe behavioral school (Cyert & March, 1963;March & Simon, 1958), a departure from stan-dard neoclassical thinking, which assumes con-stant investment in search and exploration, in-dependent of the current task. The satisficing

behavior assumption has been widely tested inand supported by management studies over thedecades (see Greve, 2003, for a recent review),and by evidence on the impact of past perfor-mance on investment in attention and learningby firms (Bateman & Zeithmal, 1989; Simon,Houghton, & Aquino, 2000; Thomas, Clark, &Gioia, 1993).

In addition, the perception of uncertainty hasbeen proposed as an important factor affectingthe evaluation of a given situation (McClure,Gilzenrat, & Cohen, 2006; Yu & Dayan, 2005).It seems that our brains distinguish betweenexpected and unexpected uncertainty. Two im-portant neuromodulatorsacetylcholine (ACh)and norepinephrine (NE), respectivelysignalexpected and unexpected sources of uncer-tainty. When exploiting, if prediction errors arehigher than expected, the current strategyshould be revised and we should explore. If, onthe contrary, the prediction errors can be ac-counted for in terms of expected uncertainty,the exploiting strategy should be maintained. In

general, taking into account that individuals arerisk-averse, then the higher the unexpected un-certainty perceived in a situation or a problem,



8/21

the more difficult it will be for the individual tounderstand the outcome of the situation/problem and so the less the utility from thesituation.

The neuroscientific findings summarizedabove provide the basis for an understanding ofwhat underlies the explorationexploitation di-lemma. These findings also help to bridge someof the gaps described in Sections 2 and 3. Theneuroscientific definitions of exploration andexploitation, and the discovery of the neurologicalmechanisms underlying the shift between the twoattentional states, help to resolve the debate overwhether exploration and exploitation are to beviewed as positions on a continuum or as orthog-onal situations. At the individual level of analysis,they are clearly orthogonal, since at any givenmoment an individual cannot be in both the phasicand tonic modes of LC functioning. At the orga-nizational level, however, members of a groupof individuals might be functioning in differentmodes, which means that the group overall willbe working on a continuum between a com-pletely phasic mode (all focused on the currenttask, i.e., purely exploitative mode) and a com-pletely tonic mode (all focused on exploration).Of course, any group typically works at a posi-

tion on the continuum located somewhere be-tween the two extremes, but this position is

important to the prediction of collective behav-ior (see Figure 2). It is important that the (dis-crete) shifts in LC operating modes of individ-uals in the group over time will cause the group

position on the continuum to constantly shift.In the next section, we offer a managerial

illustration of how these neurological mecha-nisms might influence the activities of groupsinvolved in research and development (R&D)work.

ExplorationExploitation and the Wizard

of Menlo Park

How are the ideas presented in this paper re-flected in a real case? In this section, we argue thatan organizational process of strategic relevance,such as a product innovation process, can be de-composed in subactivities (explorativeexploit-ative behaviors) that are largely influenced by adecision-makers attention focus (broad and nar-row), which we expect to in turn originate fromneuromodulatory mechanisms guided by the LC.

We will exemplify the ideas we propose us-ing the well-documented case of Thomas AlvaEdisonone of the most famous inventors ofall timesand the main events in one year

of work on the microphone at Menlo Park, oneof the first facilities entirely dedicated to R&D

If the key decision-makersattention mode is:

phasic tonic

X

phasic tonic

X

phasic tonic

X

phasic tonic

X

phasic tonic

X

and the organizational

behavior

is more likely to be:

Exploitative ExplorativeX

exploitative explorative

X

X

X

X

X





their behavior will

more likely be:

If the key decision-makers

attention mode is:

phasic tonic

X

phasic tonic

X

phasic tonic

X

phasic tonic

X

phasic tonic

X

and the organizational

behavior

is more likely to be:

Exploitative ExplorativeX


X

X

X

X

X





their behavior will

more likely be:

Figure 2. Key decision-makers attention and organizational behavior.



9/21

activities. Edison serves as a good case forillustrating the ideas presented in this paperrelating to micro- and macrolevels of analysissince he was not the lone genius but instead a

collective noun and means the work of manymen (Lindgren, 1979, p. 17 cited in Swedberg,1993). While Edison was clearly at the head ofthe Menlo Park operations, they were the resultof the collaboration and work of many individ-uals: the Park apparatus served to promote ahuge number of inventions (a minor inventionevery 10 days and a big thing every six monthsor so [Lindgren, 1979, p. 17 cited in Swedberg,1993]), generating more than 400 patents inits 6 years of operation. There is agreement on

the fact that Edison was a relentless innovatoror, closer to our argument, a meta inventor.He made Menlo Park the cornerstone of modernindustrial research. The Park was the first in-dustrial laboratory concerned with both creatingknowledge and controlling new knowledge ap-plication. It is interesting to apply the ideas wehave integrated so far to the context of an R&Dlab, where not only research was done andmany new inventions were discovered, but alsothe development, polishing, protection, and sell-ing of the inventions was done.

Menlo Park demonstrated the ability to suc-cessfully combine both explorative and exploit-ative activities. Different inventors insideMenlo Park were involved in both explorativeand exploitative activities, with Edison leadingtheir work in both cases. Edisons high involve-ment in the different activities in Menlo Parkmay have had some negative consequences forthe development of the incubator. On the otherhand, his role exemplifies how the way thatleaders attend to the particular problems of the

innovation process, facing the different deci-sions required to handle both explorative andexploitative challenges, contributes to the suc-cess of the overall innovation process.

Not only was Edison working on many dif-ferent projects at once, requiring differentmodes of operation in his attention system, buthe also showed the ability to continuously shiftbetween the two cognitive modes within thecontext of a single process. On the one hand hecould be very focused on advancing knowledge

development toward the solution of a very spe-cific problem. On the other hand, he also recom-bined specialized knowledge from different

fields to generate new, broader subfields andmore general knowledge.

How did he switch from one mode to theother? In different innovation cases there is

evidence of the importance of specialization asdomain-specific expertise (Kaufman & Baer,2006). Specific expertise is usually the founda-tion for building innovation. In addition, how-ever, in Edisons case (as in other cases studiedby Kaufman and Baer), general expertise playeda key role, along with the ability to recombineknowledge from different fields, or to apply itfrom one field to another. As stated by Harga-don and Sutton (2000), Edison and the people inhis lab had the ability to move easily in and outof separate pools of knowledge, to keep learn-ing new ideas, and to use ideas in novel situa-tions (p. 161). Like Alexander Graham Bell,Samuel Morse, Henry Ford, and others, Edisondid not advance science in the way specialistsdo. These scientists instead focused on and de-veloped an in-depth knowledge about the spe-cific issues (problems, components, etc.) oftheir inventions, while also broadening andbringing different streams of scientific discov-eries into practical devices and systems (Skra-bec, 2006). These serial innovators were some-

times able to explore, broaden their attention, becreative, recombine and bring together differentideas and also exploit, narrow down their atten-tion, concentrate, and focus on solving a spe-cific problem.

To illustrate, a detailed study by Carlson andcolleagues (Carlson, 2003, p. 155156) showshow Edisons work on the microphone (thecarbon button transmitter for the telephone) canbe summarized as a series of contrasting behav-iors in which During certain periods, Edisonvaried his lines of research, and then at partic-ular moments, he appears to have selected oneline for further development (p.156). His in-ventive patterns can be seen as characterized byexplorative periods (when he played out dif-ferent lines of investigation) following by ex-ploitative periods when he selected and fo-cused by either choosing one of the lines or byrecombining the most promising lines.

Figure 3 summarizes the main activities Edi-son undertook while developing the micro-phone for the telephone during 1877. The ovals

show the moments when he chose which line toproceed with (Diagram adapted from Carlson,2003, p. 156). For example, in late April 1877,



10/21

Edison was experimenting with different lines

(i.e., dragging, rubbing, etc.). He then chose tofocus on one option (i.e., the pressure tele-phone) to further improve it. Again, in Septem-ber of that year, after having experimented withdifferent production models (i.e., resonance,reed, etc.), he decided to focus on one option. Inthis case he did not actually chose one, insteadcombining the most promising results from dif-ferent lines and focusing on developing a rubbertube production version.

As documented in Carlson (2003, p. 152) thisparticular situation can be analyzed as a specificcase in point:

Edison or his associate James Adams sub-stituted points of plumbago (i.e., graphite) forthe disks on his squeeze telephone. These tele-phones seemed to work better than previousversions, leading Edison to think more carefullyabout using points. In particular, he now con-sidered using four high-resistance points press-ing on the diaphragm with varying degrees offorce. Edison noted an inverse relationship be-tween the mechanical force and the electrical

resistance that the resistance increased as theforce decreased- and drew on this observation toconstruct a pressure telephone in April 1877.

A schema for interpreting this example in the

light of the ideas proposed in this paper ispresented in the Figure 4. The concentric ovalssignal the different sublevels at which theexplorationexploitation dilemma can be seenin the telephone example. At the center wepresent the more micro level we have includedin our ideas (the LC mode), and the most ex-ternal oval represents the more macro level wehave presented (the behavior at the organiza-tional level). In the telephone example, Edisonstarted with an exploratory behavior searching

for alternatives (in this case five different ones)and experimenting (adding the graphite points).Edison got positive feedback from the environ-ment, obtaining a good outcome (better workingtelephones), which increased the utility andlowered the uncertainty he perceived in theproblem at hand. Perceiving a higher utility, thefocused mode of LC functioning (phasic) arose(see first central box to the left). Edison startedfocusing his attention (think more carefullyabout using points. . . p.156) on one alternative

to improve it. He narrowed down from fivealternatives to the apparently best one to furtherexploit the idea, develop and improve it (he

Jan Feb Mar Apr Jun Jul Aug Sep Oct Nov Dec

Choosing a type of t elephone

Further refinement and polishing of the design to be patented

Choosing an appropriate material for a production model

Dragging

Rubbing

Switching

Induction

Squeezing Pressure

Resonance

Reeds

Springs

Fluff

Rubber

Diaphragm

Magneto

Battery

Transformer

Combin. Transmitter

Induction

Felt

Battery telephonepatented

The diagram summarizes the lines of research Edisonundertook during 1877.The ovals show the moments when he chose among

those lines which to proceed with (Diagram adaptedfrom Carlson, 2003, p.156).

Figure 3. Edisons lines of research for developing the telephone.



11/21

chose to drop four lines in favor of the pressureline p. 156).

This shifting between exploring and exploit-ing was repeated several times in the telephoneinvention and in several others. As can be seenin Figure 4, as a result of the squeezing alter-native, Edison focused his research on the pres-sure telephone. His shifts between exploringand exploiting continued during the summer of1877 (in Figure 4 see the box beginning withnot convinced). After deciding to focus on thepressure telephone, Edison and his associatesfaced a different problem: the quality of theacoustic components was not satisfactory forthem (low utility perceived). Consequently, hedecided to undertake a search for better compo-nents (perceiving a low utility, he was under thetonic LC mode and so broadened his attention).The lab then extended their investigation todifferent acoustic components (material search,diaphragm, resonance cavity, reeds, springs,

fluff) that were studied and tested (explorativebehaviors). Once higher standards were reached(and so a high utility perceived) again a narrow-

ing-down process appeared in which Edisoncombined the most promising results from thedifferent lines of research to create the rubbertube production version of the telephone.

If we think about applying the schema inFigure 2 to interpretsomewhat liberallyEdisons case, we can see that if we take a staticpicture at a moment when Edison and his asso-ciates were exploring different lines, we maysee how their broad attention (resulting from atonic LC mode) reflected an explorative behav-ior at the individual level and aggregated also atthe organizational level. At other moments (as,e.g., when narrowing down to one optionamong the many explored) the more focusedattention (resulting from a phasic LC mode)showed exploitative behaviors.

It is important, for our argument, there is noevidence that Edison relied on different peopleto allocate exploration- or exploitation-orientedtasks, but that all the people involved in his labs

tackled both types of challenges. Having indi-viduals capable of shifting easily from exploi-tation to exploration and back also allows a

LC mode

broadtonic

Attention mode

phasic

Picking one of the research

lines: the pressure telephone

narrow explorative

Behavior at

individual level

exploitative

Experimenting with 5

different lines of research

Behavior at

organizational level

drop four lines

in favour of the

pressure line

undertook a brief search for

an alternative

undertaking intensive studiesof the acoustic

made a series of changes

and they tested several


not convinced

dissatisfied with

the volume andarticulation of

these

telephones

think more

carefully aboutusing points

Figure 4. Example: Edison and the telephone invention.



12/21

group to do so: aggregated at the organizationallevel, the combination of explorative and ex-ploitative behaviors is more likely to lead to anadaptive behavior that swiftly moves between

the exploring

exploiting modes.

The Challenge of Measurement and

Empirical Validation

In applying neuroscientific findings and tech-niques to management problems, we need to beaware of some key epistemological differencesbetween these research programs. Neuroscien-tists, and neuroeconomists for the large part, areinterested in the average effects of their exper-imental manipulations in identifying the neuralcorrelates of a given process (e.g., a utilityevaluation) triggered by a homogeneous stimu-lus (e.g., increased levels of uncertainty) andproducing a set of possible, observable, deci-sional outcomes (e.g., a type of decision in agame-theoretic scenario). Consequently, thepursuit of these types of research questions andthe experimental designs generated tend to ne-glect individual variations around the mean re-sponse to the stimulus, and in fact considerthem a nuisance (Frederick, 2005), since the

larger the variance the weaker are the mean-based results.

However, as Lubinsk1 and Humphreys(1997) note, a neglected aspect does not disap-pear because it is neglected, and there is nogood reason for ignoring the individual varia-tion around a mean response, especially if wehave good logic to expect an important causallink between managerial decision-making andorganizational performance, or, as in this paper,between the allocation of attention as an ante-cedent of decision-making behavior and its con-sequences in organizational behavior and per-formance. As neuroscientists do, we should careabout what causes the average to work in acertain way, but our focus must be on theexplanation of the differences: what mightlead certain individuals and their organiza-tions to display diverse reactions to similarstimuli, and consequently different levels ofperformance? This basic difference in episte-mological approaches implies some chal-lenges to management researchers who need

to adjust some of the traditionally adoptedneuroscience research methods in order tocomply with their objectives.

A first challenge is that testing for differencesrequires a larger sample and implies highercosts in terms of time, laboratory access, anddata processing. Also, the type of participants is

different: many applications of neuroscience toeconomics, marketing, and finance use datafrom college students. However, if we want tounderstand differences in managerial decision-making behavior, it is necessary to sample man-agers, executives, and entrepreneurs of differ-ent kinds, with the aim of tracing differencesin the neural correlates that antecede theirbehavior, and also to eventually be able tocorrelate the results on individual character-istics with those of the organizations orgroups over which they exert significant in-fluence. Differences in age and experiencemay shape the brain structure and the activa-tions that can be observed. One interestingsetting for testing our ideas would be amongthe entrepreneurs of small family enterprises.In those organizations, as described by Bod-mer and Vaughan (2009), the limited re-sources will possibly increase the impact onthe organizational moves of what the entre-preneur attends to and the cognitive maps hedoes develop. Finally, replication is espe-

cially important in managerial studies sincecontextual effects related to task features,firm and sector characteristics, cultural traitsand institutional conditions may alter the wayindividuals make decisions following thesame set of stimuli, and the way organizationsconsequently perform.

Another distinctive feature of managementscholarship requires researchers to go beyonduncovering what the effects of certain abilitiesmight be, and to attempt to understand howthese cognitive abilities can be developed andused successfully to improve organizational per-formance, so that abilitiessay the flexibilityto shift attentional mode as soon as unex-pected uncertainty levels changecan be dif-fused to benefit organizational outcomes. Oneof the key principles of behavioral neuro-science, in fact, is that experience can modifybrain structure long after brain developmentis complete. Brain plasticity refers preciselyto the brains ability to change structure andfunction. Experience is a major stimulant of

brain plasticity and works by producing mul-tiple, dissociable changes in the brain includ-ing increases in dendritic length, increases (or



13/21

decreases) in spine density, synapse forma-tion, increased glial activity and altered met-abolic activity (Kolb & Whishaw, 1998). Re-search on humans produced the interesting

result that the plasticity of the nervous sys-tem remains throughout the life span and ex-tends well into old age (Taub, 2004). If theantecedents to certain managerial abilities areidentified, it may imply that it might be pos-sible to modify brain structure through differ-ent types of exercises and training, and thusgain abilities relevant to improving manage-rial functioning following different experi-ences and at different ages.

To design a study for testing the neuroscien-tific findings in a managerial setting it is neces-sary to create or adapt experimental tasks thatcover a series of steps corresponding to thedifferent constructs illustrated by the neurosci-entific findings. An ideal option would be toobtain direct measures of the managers atten-tion focus while facing different real-life deci-sions and to correlate such measures with theperformance obtained out of those decisions(both at a purely individual level and at anorganizational level) both in the short- and thelong-term. However, given the difficulty (or

impossibility) of measuring managers in thereal context with the actual brain imagingtools, a good viable proxy would be to corre-late the performance obtained, for example, inmanagerial decision-making simulations fac-ing the explorationexploitation dilemmawith the precise attention focus measures ob-tained while completing different neuropsy-chological tasks in a lab context. There aredifferent alternatives that researchers couldfollow to measure the ability of making deci-sions regarding exploring exploiting, someof which are also compatible with brain im-aging techniques that could allow researchersto measure not only performance in theexplorationexploitation decisions (the ob-served behavior) but also the neural correlatesof such decisions.

We now turn to discuss how each of the keyconstructs in this paper could be actually ob-served with the context of one specific task,which can be administered using fMRI tech-niques: the gambling task (Daw et al., 2006).

To understand the antecedents of the deci-sion-making ability related to managing theexplorationexploitation dilemma, we have

proposed four constructs based on the findingsin neuroscience. The first construct is the levelof uncertainty connected to a given decisionalsituation. The first step in the experimental de-

sign is thus to evaluate the managers percep-tion of the uncertainty of the outcomes in thetask. This perception of uncertainty translatesinto a utility assessment, the second con-struct. Different parts of the brain intervene inutility assessment and, as we have shown, theACC and OFC play important roles. Depend-ing on the managers assessment of utility inthe current task, an attention mode arises, thethird construct. As already explained, the LCplays a fundamental role in modulating the

attention mode according to the perceivedutility. As a consequence of the attentionmode, the manager will then show a behavior,the fourth construct. In the case of broadattention (LC active in tonic mode), explor-ative behaviors arise and managers will act(or will propose solutions) in ways character-ized by experimentation, flexibility, discov-ery, and innovation. In the case of focusedattention (LC active in phasic mode), exploit-ative behaviors will occur and managers will

act selectively, according to refinement ofcurrent processes and efficiency in the currenttask. If the behavior matches what the situa-tion demand (e.g., high uncertainty matchedwith explorative behavior and low uncertaintywith exploitative behavior), higher perfor-mance can be expected.

It is important that the ability to balanceexplorationexploitation through flexible man-agement of the situational requirements toachieve the appropriate attentional response can

only be assessed if the decision process is rep-licated under stable contextual conditions. Thisadaptive process at the individual level can belinked, and the link empirically tested, to theorganizational ability to balance the explora-tionexploitation dilemma.

The four constructs we propose can be mea-sured in an experimental context. These exper-iments involve two types of data. First, theyrequire data derived from behavioral measure-ments during experimental tasks (e.g., response

times, performance, etc.) and from question-naires and interviews. Second, they require dataon brain functioning, which can be obtained



14/21

during experiments by using brain imagingtechniques.2

The particular technique to be used can beselected depending on the properties of

brain

behavior association to be observed. Ifthe study requires a high spatial resolution,fMRI and PET will be required. If the studyrequires high temporal resolution, EEG or MEGwould be suitable. A study that requires bothhigh temporal and high spatial resolution coulduse a combination of these techniques (such asfMRI or PET with EEG/MEG).

Gambling in the MRI Machine

An experimental task that could assess theexplorationexploitation dilemma at the indi-vidual level and is compatible with the brainimaging techniques just exposed is the gam-bling task as adapted by Daw et al. (2006). Inthis experiment, participants play by choosingamong four slot machines, to win as manypoints as possible. They are faced with theclassical explorationexploitation dilemma in achanging environment context. During the ex-periment different characteristics on the ma-chines are manipulated (payoff average, uncer-

tainty of returns, etc.) and participants mustchoose whether to continue to play on a partic-ular machine or explore new possibilities in thehope of earning more points. While the individ-ual is playing, her or his brain can be scanned(e.g., using fMRI) to obtain measures for eachof the four constructs developed above. Ofcourse, the task could be done in a normal PCand obtain only the behavioral performancemeasures of explorationexploitation decision-making, depending on the strategies and payoffsobtained by the participant. Also important, thesoftware controlling the game allows a goodlevel of manipulation of the uncertainty level. Inaddition, the utility perception and the attentionmode can be measured using brain imagingtechniques. Most management studies proxy forattention by time allocated to an activity (e.g.,the famous study by Mintzberg, 1973, whoseresults were confirmed by Kurke and Aldrich,1983). Other studies proxy for attention by thenumber of times a person refers to a certainissue (e.g., the number of sentences in a share-

holder letter referring to a specific element(DAveni and MacMillan, 1990; Barr, Stimpert,& Huff, 1992). However, the ideas we propose

do not require a measure for general attention;rather, we want to differentiate between twotypes of attentionbroad and focused. The at-tention mode can be assessed by observing LC

functioning using a brain imaging technique(e.g., the participants can play the task whilelying in the MRI scan) or by measuring pupildiameter, found to correlate remarkably wellwith LC tonic activity (Gilzenrat, Cohen,Rajkowski, & Aston-Jones, 2003). The diame-ter of an individuals pupils changes under var-ious conditions. For instance, the pupil diameterof someone who is thinking increases; the pupilof a tired person shrinks. These effects havebeen proven through a number of psychology

(Shinoda & Kato, 2006) and neuroscience ex-periments (Gilzenrat et al., 2003).

Finally, behavior can be observed based onindividual choices. In this way, one task could beused to measure the decision-making performancewhen facing the explorationexploitationtradeoff, and for manipulating and measuring thedifferent constructs that we have illustrated andthat affect such performance (uncertainty, per-ceived utility, attention focus). Table 2 summa-rizes how each construct can be defined and mea-

sured in an experiment such as the gambling task.This gambling task could be complemented

with other alternatives, such as simulation, de-cision-making vignettes and self-reportedscales, to have a more reliable measure. A treat-ment of all the available alternatives using thesetechniques is out of the scope of this article, butthe authors will be pleased to provide a synopsisand an assessment of results from ongoing em-pirical work to all interested scholars.

2 The basic set of techniques used to generate neurolog-ical images is electroencephalography (EEG), magneto en-cephalography (MEG), positron emission tomography(PET), and fMRI. There are many limitations to the use ofthese techniques: they are expensive to operate and resultsare open to subjective interpretation. They are also intrusivefor subjects although to different degrees. EEG and MEGare considered the least intrusive, while fMRI may causemany subjects discomfort related to having to lie still in asmall space. Researchers must take the degree of intrusioninto account since it affects data gathering, particularly if

the subjects are busy managers and key decision-makers.Nevertheless, these techniques offer by far the best physicalevidence to date on the activation of specific parts of thebrain consequent to given stimuli.



15/21

For the purpose of testing the ideas proposedin this manuscript,3 the first best option is cer-tainly to correlate the direct measures of themanagers attention focus in different real-lifesituations with the performance obtained in thereal decisions (both at a purely individual level

and at an organizational level) both in short- andlong-term. However, given the difficulty (orimpossibility) of measuring managers in realsituations with the actual brain imaging tools, agood viable proxy would be to correlate atten-tion focus measures obtained while performingdifferent neuropsychological tasks in a labcontext, with for example, managerial decision-making simulations facing the explorationexploitation dilemma. The performance mea-sures obtained in these simulations can becompared with the measures obtained in self-reported scales such as the cognitive flexibilityscale (Martin & Rubin, 1995) or the recent scaleof explorationexploitation activities by Momet al. (2009).

Conclusions

This paper addresses a dilemma common toorganizations. Managing the trade-off betweenexploitation and exploration is fundamental toadaptive behavior and learning in increasingly

complex and rapidly changing contexts. Al-though there has been much research on thistrade-off, there are still several key gaps in the

literature. First, we still know little about howexploration and exploitation are actually done.Second, the appropriate level of analysis atwhich the explorationexploitation tradeoff issolved is not clear. Third, in many cases what ismeant by explorationexploitation is not very

clear. Fourth, variation among individuals in thetendency to respond in an exploitative or ex-plorative way to a given stimulus, and on theability to shift their responses according tochanges in the environmental conditions, havenot been explored. To start addressing these keyanalytical gaps, we thus propose a frameworkand a method which, in our view, contributesparticularly to the microlevel problem of indi-vidual-level variance in behavior. We provide adefinition that is compatible with the manage-

ment and the neuroscientific literature that weare using. We focus on the individual as thefundamental unit of analysis and study howexploration and exploitation are done at themicro, neurological level in terms of the pro-cesses going on in an individuals mind and theensuing behavior when faced with a given en-vironmental stimulus. Why does this frameworkmatter? We believe there are at least four pos-

3

While at the time of writing the study was still in itsinfancy, at the time of publishing the ideas were imple-mented along lines consistent with those here discussed andvery promising preliminary results started to be obtained.

Table 2 Example of How Constructs Could Be Defined and Measured in a Multi-Armed Gambling Task

Construct Definition Possible measure

Situation uncertainty It can be defined in terms of the following: 1. Frequency (#

times task is repeated during x time). 2. Heterogeneity(degree of novelty); 3. Causal ambiguity (number anddegree of interdependence of subtasks; degree ofsimultaneity among subtasks) (Zollo and Winter, 2002)

Payoffs change randomly from trial

to trial (manipulation of differentsituation characteristics)

Perceived utility How much the individual likes the possible outcomes ofthe present situation (taking into account theanticipation of current outcomes and the memory ofpast decisions) (Aston-Jones and Cohen, 2005)(highlow on a scale depending on method)

ACC and OFC functioning usingfMRI

Attention mode What type of attention is devoted to a situation: broad ornarrow (Aston-Jones and Cohen, 2005)

LC operation mode using fMRI orproxy with pupil diameter

Behavior How much search for new alternatives is done (highbeing exploration, low exploitation) (Aston-Jones andCohen 2005)

The observed strategy followed bythe individual (i.e. her/his choicein each trial)

Note. ACC anterior cingulate cortex; OFC orbitofrontal cortex; fMRI functional Magnetic Resonance Imaging;LC locus coeruleus.



16/21

sible areas of contribution for the ideas putforward in this paper.

How the Tradeoff Is Solveda Micro

Perspective

As discussed earlier in this paper, there is stilla remarkable lack of clarity as to what is theappropriate level of analysis for understandingthe trade-off between exploitation and explora-tion. At the organizational level, many havefocused on the attributes that make organiza-tions more or less explorative (e.g., Tushman &OReilly, 1996). Others have argued instead(e.g., Papo, 2007) that individual-level pro-cesses ought to play a more central role inunderstanding the origin of new ideas. Obvi-ously, these issues are related. Yet, we stillknow very little about how microlevel pro-cesses build up to organizational outcomes. Weused the example of Menlo Parks activities toargue that the crucial issue here is not that ofallocating exploitative versus explorative tasksto those individuals or organizations which arebest suited to that task. Rather, our frameworkpoints to the idea that what matters to balancethe tradeoff is not specialization, but the ability

of key decision-makers to shift seamlessly fromone task to the other, as we illustrated with theprocess Edison enacted in one of his key inno-vations. At the organizational level, this argu-ment is consistent with recent research on thedynamics of innovation in complex technicalsystems, which has warned against the dangersof strategic outsourcing, often grounded in thebelief that exploitation and exploration are ac-tually separate activities which can be attributedto different types of organizations (e.g., Bru-soni, Prencipe, & Pavitt, 2001). While there issome evidence of the dangers of outsourcingstemming from excessive specialization, micro-founded explanations are in short supply. Theapplication of neuroscience to managerial deci-sion-making offers important opportunities forthis line of inquiry to be developed to includethe study of interactions among individuals en-gaged in these high-level cognitive tasks, andthe generation of collective results that go be-yond the sum of individuals capacities. Whilethe aggregation of individual capacities and

behaviors into organizational ones remains amain limitation from the neuroeconomics field(Papo, 2007), this paper attempts, at least, to

identify similarities between micro- and organi-zation-level processes which ought to be furtherexplored to generate sensible microfoundationsof macrobehaviors.

How the Tradeoff Is Solveda Macro

Perspective

Domains that have traditionally focused onmacro-organizational analyses such as organi-zation theory, strategy, and entrepreneurshipmight benefit from the development of a theoryof exploration and exploitation based on theneurological processes occurring within indi-vidual managers brains.

For example, consider one of the basic in-sights in Marchs (1991) seminal article, whichis related to the fact that exploration requiresheterogeneity in human capital, which disap-pears without constant personnel turnover be-cause the newcomers learn the code and adaptto the firms modus operandi. March concludes(somewhat paradoxically) that the slower thelearning of the code, the higher the explorationin the firm. However, the analysis in this paperpoints to an alternative mechanism which doesnot rely on personnel turnover to balance the

explorationexploitation dilemma, but ratherfocuses on the organizational members neuro-logical characteristics, that is, their propensityto function in tonic or phasic mode in the neu-romodulation of their attention. It could bespeculated that a broad attention span caused bythe tonic activity of LC leads to a relativelyslower socialization process (i.e., learning of thecode) and thus more exploration at organiza-tional level.

For scholars engaged in the study of entre-preneurship, our work offers the possibility totheorize and empirically validate (with objec-tive measures) insights related to the explana-tion of explorative decisions, which might gen-erate the foundation of new ventures internal tothe firm (entrepreneurship) or outside it (spin-offs or start-ups). The launch of a novel enter-prise can be viewed as the consequence of thecontinuous allocation of attention to explor-ative processes, and the development of neu-rological foundations for such choices can beparticularly useful to this field. For similar

reasons, strategy scholars might be able tobuild on a better understanding of the neuro-logical foundations of explorationexploita-



17/21

tion decisions to develop models of strategicchoice based on the idiosyncratic characteris-tics of the strategy-makers, over and abovethe influence of the industry context and the

organizational endowment.

Enhancing Attention Modulation

Third, and more practically, the frameworkwe propose in this paper might be used to de-velop teaching tools to foster individuals abil-ity to control and properly shift the focus oftheir attention. There is an element of plasticityin the neurological processes we discuss here,which can be leveraged to develop training pro-

grams to improve individuals responses tochanging environmental circumstances. Sinceattention modulation affects the ability to prop-erly make decisions regarding explorationexploitation, it will be then possible to adaptsome instruments which have been developedfor improving the attentional control (Sohlberg& Mateer, 1989, 2001). Very interestingly, forexample, the analysis of the modulation of at-tention and its impact on decision-makingmight open new venues in managerial andentrepreneurial education through the develop-ment of tools and techniques which manage-ment students and practitioners can use to en-hance their attentional control. These tools andtechniques can be very varied, ranging fromtraditional in-class methods (e.g., for acquiringawareness of perceptive biases) to mental con-trol practices (as, e.g., neuroimaging studieshave demonstrated the positive impact of med-itative practices on the improvement of func-tional up-regulation of brain regions affectingattention control. For a recent review see Rubia,

2009). In addition, since many psychiatric dis-orders of higher level cognition are thought tobe due to deficits of attention (Posner & Pe-tersen, 1990) researchers could adjust the toolsused to treat patients of attention disorders to beapplied to managers aiming to improve theirattention control.

Reflection and Some Convergence Among

Different Domains

On a broader note, a fourth contribution aimsat reaching out to the now-broad crowd in socialsciences interested in exploring the neuroscien-

tific foundations of economic and organiza-tional behavior. Social sciences are highly seg-mented in different knowledge domains. Thefindings of our work might foster a more inte-

grated discussion. For example, mainstreameconomists might consider it interesting tomove beyond the replication of results based ongame theory, the focus of most neuroeconomicswork so far, to develop novel insights into theneurological foundations of economic behavior.Also, the shift toward the explanation of vari-ance across individual traits and consequent be-havior, as opposed to the characterization of themean tendency for the population of decision-makers, might serve to correct a standard bias inthe discipline.

On the other side of the disciplines fence,evolutionary economists will appreciate our at-tempt to develop some of the microfoundationsof the work related to how firms learn andevolve. The appropriate balancing of explor-ative and exploitative search is arguably a cor-nerstone of their theories. Depending on theattention mode, a certain way to decompose theproblem will emerge and the search space willbe defined accordingly. Search, therefore, is aconsequence of the attention mode. If the atten-

tion focus is broad, for example, problems willbe decomposed in broader modules, while anarrow attention focus will result in problemsbeing decomposed in finer modules. Brusoni,Marengo, Prencipe, and Valente (2007) showthat the balance to this trade-off depends uponthe volatility of the problem environment. Instationary environments there is an evolutionaryadvantage to overmodularization, while inhighly volatile environments, contrary to thereceived wisdom, modular search is inefficientin the long run. Similarly, we propose that inlow-uncertainty tasks, exploitative behavior(i.e., deriving from a narrow attention mode andinvolving high decomposition of problems)would be advantageous, while the opposite willbe true in the case of highly uncertain tasks:high decomposition is not efficient and explor-ative behavior derived from broad search willlead to better performance.

The organizational behavior field, with itsorientation toward studying individual behaviorin firms, is the obvious audience for the ideas

discussed here. The main value for scholars inthis field is that our findings expand their re-search agenda, which traditionally has been



18/21


19/21

and convergent experimental findings. StrategicManagement Journal, 10, 5974.

Beckman, C. M. (2006). The influence of foundingteam company affiliations on firm behavior. Acad-emy of Management Journal, 49, 741758.

Beversdorf, D. Q., White, D. M., Chever, D. C.,Hughes, J. D., & Bornstein, R. A. (2002). CentralB-adrenergic modulation of cognitive flexibility.Cognitive Neuroscience and Neuropsychology, 13,

25052507.Bhatt, M., & Camerer, C. F. (2005). Self-referential

thinking and equilibrium as states of mind ingames: FMRI evidence. Games and Economic Be-havior, 52, 424459.

Bodmer, U., & Vaughan, D. R. (2009). Approachesto preventing crises in family controlled small en-terprises. Journal of Neuroscience, Psychology,

and Economics, 2, 4158.Botvinick, M. M., Cohen, J. D., & Carter, C. S.

(2004). Conflict monitoring and anterior cingulatecortex: An update. Trends in Cognitive Sci-ences, 8, 539546.

Brusoni, S., Luigi, M., Prencipe, A., & Valente, M.(2007). The value and costs of modularity: A prob-lem-solving perspective. European ManagementReview, 4, 1539.

Brusoni, S., Prencipe, A., & Pavitt, K. (2001).Knowledge specialisation, organizational couplingand the boundaries of the firm: Why firms knowmore than they make? Administrative ScienceQuarterly, 46, 597621.

Carlson, W. B. (2003). Invention and evolution: Thecase of Edisons sketches of the telephone. In J.Ziman (Ed.), Technological innovation as an evo-lutionary process. Cambridge, UK: CambridgeUniversity Press.

Carstensen, L. L., Isaacowitz, D., & Charles, S. T.(1999). Taking time seriously: A theory of socio-emotional selectivity. American Psychology, 54,165181.

Cohen, J. D., Aston-Jones, G., & Gilzenrat, M. S.(2004). A systems-level perspective on attention

and cognitive control, guided activation, adaptivegating, conflict monitoring, and exploitation versusexploration. In M. I. Posner (Ed.), Cognitive neu-roscience of attention (pp. 7190). New York:Guilford Press Publications.

Cohen, J. D., McClure, S. M., & Yu, A. (2007).Should I stay or should I go? How the human brainmanages the trade-off between exploitation andexploration. Philosophical Transactions of theRoyal Society, 362, 933942.

Cyert, R. M., & March, J. G. (1963). A behavioraltheory of the firm. Englewood Cliffs, NJ: PrenticeHall.

DAveni, R. A., & McMillan, I. C. (1990). Crisis andthe content of managerial communications: Astudy of the focus of top managers in surviving and

failing firms. Administrative Science Quar-terly, 35, 634657.

Daw, N. D., ODoherty, J. P., Dayan, P., Seymour,B., & Dolan, R. J. (2006). Cortical substrates forexploratory decisions in humans. Nature, 441,876879.

Eisenberger, N. I., Lieberman, M. D., & Williams,K. D. (2003). Does rejection hurt? An fMRI studyof social exclusion. Science, 302, 290292.

Felin, T., & Foss, N. (2005). Strategic organization:A field in search of micro-foundations. StrategicOrganization, 3, 441455.

Felin, T., & Hesterly, W. S. (2007). The knowledge-based view, nested heterogeneity, and new valuecreation: Philosophical considerations on the locusof knowledge. Academy of Management Re-view, 32, 172202.

Frederick, S. (2005). Cognitive reflection and deci-sion making. Journal of Economic Perspec-tives, 19, 2542.

Gibson, C. B., & Birkinshaw, J. (2004). The anteced-ents, consequences and mediating role of organi-zational ambidexterity. Academy of ManagementJournal, 47, 209226.

Gilzenrat, M. S., Cohen, J. D., Rajkowski, J., &Aston-Jones, G. (2003). Pupil dynamics predictchanges in task engagement mediated by locuscoeruleus. [Abstract] Society for Neuroscience Ab-stracts No. 515.19.

Glimcher, P. W., & Rustichini, A. (2004). Neuroeco-nomics: The concilience of brain and decision.Science, 306, 447452.

Greve, H. R. (2003). Organizational learning from performance feedback: A behavioral perspective

on innovation and change. Cambridge, UK: Cam-bridge University Press.

Gupta, A. K., Smith, K. G., & Shalley, C. E. (2006).The interplay between exploration and exploita-tion. Academy of Management Journal, 49, 693706.

Hargadon, A., & Sutton, R. I. (2000). Building andinnovation factory. Harvard Business Review,

MayJune, 157166.Holmqvist, M. (2004). Experiential learning pro-

cesses of exploitation and exploration within andbetween organizations: An empirical study ofproduct development. Organization Science, 15,7081.

Kaufman, J. C., & Baer, J. (2006). Creativity andreason in cognitive development. Cambridge, UK:Cambridge University Press.

Knutson, B., Fong, G. W., Bennett, S. M., Adams,C. M., & Hommer, D. (2003). A region of mesialprefrontal cortex tracks monetarily rewarding out-comes: Characterization with rapid event relatedfMRI. NeuroImage, 18, 263272.

Koepp, M. J., Gunn, R. N., Lawrence, A. D., Cun-ningham, V. J., Dagher, A., Jones, T., . . . Grasby,



20/21

P. M. (1998). Evidence for striatal dopamine re-lease during a video game. Nature, 393, 266268.

Kolb, B., & Whishaw, I. Q. (1998). Brain plasticityand behavior. Annual Review of Psychology, 49,4354.

Kurke, L. B., & Aldrich, H. E. (1983). Mintzberg wasright! A replication and extension of the nature ofmanagerial work. Management Science, 29, 975984.

Leonard-Barton, D. (1995). Wellsprings of knowl-edge. Cambridge, MA: Harvard Business SchoolPress.

Levinthal, D. A., & March, J. G. (1993). The Myopiaof learning. Strategic Management Journal, 14,95112.

Levinthal, D. A., & Rerup, C. (2006). Crossing anapparent chasm: bridging mindful and less-

mindful perspectives on organizational learning.Organization Science, 17, 502513.

Lubatkin, M. H., Simsek, Z., Ling, Y., & Veiga, J. F.(2006). Ambidexterity and performance in small-to medium-sized firms: The pivotal role of topmanagement team behavioral integration. Journalof Management, 32, 646672.

Lubinski, D., & Humphreys, L. (1997). Intelligence. Incorporating General Intelligence into Epidemi-

ology and the Social Sciences, 24, 159201.March, J. G. (1991). Exploration and exploitation in

organizational learning. Organization Science, 2,7187.

March, J. G. (2006). Rationality, foolishness andadaptive intelligence. Strategic Management Jour-nal, 27, 201214.

March, J. G., & Simon, H. A. (1958). Organizations.New York: Wiley.

Martin, M. M., & Rubin, R. B. (1995). A new mea-sure of cognitive flexibility. Psychological Re-ports, 76, 623626.

McClure, S. M., Gilzenrat, M. S., & Cohen, J. D.(2006). An explorationexploitation model basedon norepinephrine and dopamine activity. In Y.Weiss, B. Sholkopf, & J. Platt (Eds.), Advances in

neural information processing systems (Vol. 18,pp. 867874). Cambridge, MA: MIT Press.

McClure, S. M., Laibson, D. I., Loewenstein, G., &Cohen, J. D. (2004). Separate neural systems valueimmediate and delayed monetary rewards. Sci-ence, 306, 503507.

McKenzie, J., Woolf, N., van Winkelen, C., &Morgan, C. (2009). Cognition in strategic deci-sion-making: A model of nonconventional think-ing capacities for complex situations. ManagementDecision, 47, 209232.

Mintzberg, H. (1973). The nature of managerialwork. New York: Harper & Row.

Mom, T. J. M., Van Den Bosch, F. A. J., & Volberda,H. W. (2007). Investigating managers explorationand exploitation activities: the influence of top-

down, bottom-up, and horizontal knowledge in-flows. Journal of Management Studies, 44, 910931.

Mom, T. J. M., Van Den Bosch, F. A. J., & Volberda,H. W. (2009). Understanding variation in manag-ers ambidexterity; investigating direct and inter-action effects of formal structural and personalcoordination mechanisms. Organization Sci-ence, 20, 812828.

Montague, P. R., King-Casas, B., & Cohen, J. D.(2006). Imaging valuation models in humanchoice. Annual Review of Neuroscience, 29, 417448.

Murray, E. A., ODoherty, J. P., & Schoenbaum, G.(2007). What we know and do not know about thefunctions of the orbitofrontal cortex after 20 yearsof cross-species studies. Journal of Neuro-

science, 27, 81668169.Nelson, R. R., & Winter, S. G. (1982). An evolution-

ary theory of economic change. Cambridge, MA:Harvard University Press.

OReilly, C. A., & Tushman, M. (2004, April). Theambidextrous organization. Harvard Business Re-view, 18.

Papo, D. (2007). Where do ideas come from? Brainactivity and economic systems. European Journalof Economic and Social Systems, 20, 163186.

Perretti, F., & Negro, G. (2006). Filling empty seats:How status and organizational hierarchies affectexploration and exploitation in team design. Acad-emy of Management Journal, 49, 759777.

Peyron, R., Laurent, B., & Garca-Larrea, L. (2000).Functional imaging of brain responses to pain. Areview and meta-analysis (2000). Neurophysiolo-gie Clinique/Clinical Neurophysiology, 30, 263288.

Posner, M. I., & Petersen, S. E. (1990). The attentionsystem of the human brain. Annual Review ofNeuroscience, 13, 2542.

Raisch, S., & Birkinshaw, J. (2008). Organizationalambidexterity: Antecedents, outcomes, and mod-erators. Journal of Management, 34, 375409.

Rolls, E. (2000). The brain and emotion. Oxford,UK: Oxford University Press.

Rothaermel, F. T., & Hess, A. M. (2007). Buildingdynamic capabilities: Innovation driven by indi-vidual-, firm-, and network-level effects. Organi-zation Science, 18, 898921.

Rubia, K. (2009). The neurobiology of Meditationand its clinical effectiveness in psychiatric disor-ders. Biological Psychology, 82, 111.

Sara, S. J. (2009). The locus coeruleus and noradren-ergic modulation of cognition. Nature ReviewsNeuroscience, 10, 211223.

Shinoda, T., & Kato, M. (2006, October). A pupil di-ameter measurement system for accident preventionsystems, man and cybernetics. Paper presented at theIEEE International Conference, Taipei, Taiwan.



21/21

Simon, H. A. (1985). Human nature in politics.American Political Science Review, 79, 293304.

Simon, M., Houghton, S. M., & Aquino, K. (2000).Cognitive biases, risk perception, and venture for-mation: How individuals decide to start compa-nies. Journal of Business Venturing, 15, 113134.

Skrabec, Q. R. (2006). George Westinghouse: Gentlegenius. New York: Algora Publishing.

Smith, W. K., & Tushman, M. (2005). Managingstrategic contradictions: A top management modelfor managing innovation streams. OrganizationScience, 16(5), 522536.

Sohlberg, M., & Mateer, C. (1989). Introduction tocognitive rehabilitation: Theory and practice.

New York: Guilford Press.Sohlberg, M., & Mateer, C. (2001). Cognitive reha-

bilitation: An integrative neuropsychological ap-

proach. New York: Guilford Press.Sterpenich, V., DArgembeau, A., Desseilles, M.,Balteau, E., Albouy, G., Vandewalle, G., et al.(2006). The locus ceruleus is involved in the suc-cessful retrieval of emotional memories in hu-mans. The Journal of Neuroscience, 26, 74167423.

Swedberg, R. (1993). Explorations in economicsociology. New York: Russell Sage FoundationPublications.

Taub, E. (2004). Harnessing brain plasticity throughbehavioral techniques to produce new treatmentsin neurorehabilitation. American Psychologist, 59,

692704.

Thomas, J. B., Clark, S. M., & Gioia, D. (1993).Strategy sensemaking and organizational perfor-mance: Linkages among scanning, interpretation,action, and outcomes. Academy of ManagementJournal, 36, 239270.

Tom, S. M., Fox, C. R., Trepel, C., & Poldrack, R. A.(2007). The neural basis of loss aversion in deci-sion-making under risk. Science, 315, 515518.

Tushman, M., & OReilly, C. A. (1996). Ambidex-trous organizations: Managing evolutionary andrevolutionary change. California Management Re-view, 38, 830.

Usher, M., Cohen, J. D., Servan-Schreiber, D., Ra-jkowski, J., & Aston-Jones, G. (1999). The role ofthe locus coeruleus in the regulation of cognitiveperformance. Science, 283, 549554.

Volberda, H., Baden-Fuller, C., & Van Den Bosch,

F. A. J. (2001). Mastering strategic renewal: Mo-bilizing renewal journeys in multi-unit firms. LongRange Planning, 34, 159178.

Weick, K. E., & Sutcliffe, K. M. (2006). Mindfulnessand the quality of organizational attention. Orga-nization Science, 17, 514524.

Wittmann, M., & Paulus, M. P. (2009). Temporalhorizons in decision making. Journal of Neuro-science, Psychology, and Economics, 2, 111.

Yu, A., & Dayan, P. (2005). Uncertainty, neuro-modulation and attention. Neuron, 46, 681692.

Zollo, M., & Winter, S. G. (2002). Deliberate learn-ing and the evolution of organizational capabili-

ties. Organization Science, 13, 339352.

Showcase your work in APAs newest database.

Make your tests available to other researchers and students; get wider recognition for your work.

PsycTESTS is going to be an outstanding resource for psychology, said Ronald F. Levant,PhD. I was among the first to provide some of my tests and was happy to do so. They will beavailable for others to

Documents

LAUREIRO Etal 2010 Exploration Neuro JNPE