Sasanka Sekhar Chanda

I see tremendous potential for research by the genetic algorithm model given in Professor James March’s seminal (1991) paper introducing exploration and exploitation through an organizational learning metaphor. To me, it is one of the very few algorithms where the emergence of group-level characteristics—distinct from and not aggregate of individual characteristics—can be studied, particularly in the management and organization studies disciplines. Moreover, it is possible to study order creation in far-from-equilibrium situations, thereby freeing our minds to look beyond the confining stranglehold of research where obtaining equilibrium is considered the only worthwhile endeavor. For some reason or other, March’s study could not be exactly replicated all these years, stymieing its use in research. That issue is now resolved, as can be seen in Chanda & Miller (2019) below. The MATLAB program replicating March (1991) is available in the appendix of my CV that can be downloaded from this web-page. I shall be happy to help in developing understanding and developing modifications and extensions for further research.

Research involving Genetic Algorithm (March 1991 / Holland 1975)

Below I am providing the list of papers in the order they may be read. 

(1). Chanda SS, Miller KD (2019) Replicating agent-based models: Revisiting March’s exploration-exploitation study. Strategic Organization, 17(4): 425–449 DOI: 10.1177/1476127018815295. https://journals.sagepub.com/doi/pdf/10.1177/1476127018815295

Description. In this paper we show that the graphs underlying the March (1991) study — introducing exploration and exploitation through an organizational learning metaphor — were generated using computer code that had three additional features not mentioned in the publication text. Removal of these undocumented features puts March’s theory on firmer footing and opens up the genetic algorithm platform for multi-level theory development in a wide variety of topics in organization and management studies.

What is accomplished? This paper conclusively replicates March (1991) and suggests that the appropriate form of the March (1991) model for future work extending the model is one where the assumptions existing only in March’s code—i.e. absent from the publication text—  are taken off. An entire field of theory development is opened up by virtue of the general availability of the genetic algorithm platform for developing multi-level theory.

(2). Chanda SS, McKelvey B (2020) Back to the basics: Reconciling the continuum and orthogonal conceptions of exploration and exploitation. Computational and Mathematical Organization Theory, 26(2): 175–206 DOI https://doi.org/10.1007/s10588-020-09311-y. https://rdcu.be/b4tNx

Description. In this paper we compare and contrast the continuum and orthogonal conceptions of exploration and exploitation, elaborated in March (1991). We observe that the former maps to Astley and van de Ven’s (1983) system structural view; the latter maps to their strategic choice view. Alternately, they correspond respectively to the concepts of induced and autonomous strategic behaviors discussed by Burgelman (1983). We hope this clarification shall put to rest the debates on whether exploration and exploitation are appropriately envisioned as continuum or orthogonal constructs.

What is accomplished? We show that, when exploration and exploitation are implemented as constructs at two ends of a continuum, the change in organizational outcomes upon change in the rate of exploitation is higher (than that in the orthogonal conception), and therefore easier to detect. For this reason, managers who seek greater control of the organization may prefer a continuum configuration of exploration-exploitation activities. Likewise, researchers who seek greater ease in finding data for analysis may prefer the continuum conception as well. However, organizational outcomes from an orthogonal configuration of exploration and exploitation are far superior to that from a continuum configuration. Configuring exploration and exploitation as orthogonal activities requires extending autonomy to middle and operational levels to carry out risky (innovation-related) experiments with heterogeneous knowledge from outside the firm. Therefore, company executives need to focus less on preventing wrong-doing by organizational members, and focus more on putting in place systems and processes that allow extending autonomy to middle and operational levels to carry out risky (innovation-related) experiments.

 (Astley WG, Van de Ven AH (1983) Central perspectives and debates in organization theory. Administrative Science Quarterly, 28(2):245–273.

Burgelman RA (1983) Corporate entrepreneurship and strategic management: Insights from a process study. Management Science, 29(12):1349–1364. ).

(3). Chanda SS, Ray S (2015) Optimal exploration and exploitation: The managerial intentionality perspective. Computational and Mathematical Organization Theory, 21(3): 247–273. DOI: 10.1007/s10588-015-9184-y

Description. This paper suggests a different answer to the question what is an optimal mix of exploration and exploitation. Posen and Levinthal (2012, Management Science) say that the optimal proportion is 50:50. Posen and Levinthal (2012) assume that exploration and exploitation are two ends of a continuum. They use a single-agent bandit model to derive their answer. In our paper we model exploration and exploitation as orthogonal constructs following the multi-agent formalization given in March (1991). We show that several exploration: exploitation mixes attain the same optimal outcome. Thus, managerial intentionality is feasible: managers do not have to adopt one ‘right’ mix of exploration and exploitation. Our work demonstrates Prigogine’s principle, that diversity can be a source of continued order. It further shows that a moderate rate of inflow of diverse, un-vetted knowledge helps firms combat Knightian Uncertainty.

What is accomplished? The paper establishes that managers orienting their organization towards exploitative innovation can do equally well as managers orienting their organization towards exploratory innovation. It also shows that the prescription regarding appropriate managerial action is vastly different in open systems—where the objective is to fashion orderly structures in far-from-equilibrium conditions—compared to the prescription from research that mandates reaching equilibrium in a closed system as its main purpose.

 (Posen HE, Levinthal DA (2012) Chasing a moving target: Exploitation and exploration in dynamic environments. Management Science 58(3):587–601 ).

(4). Chanda SS (2017) Inferring final organizational outcomes from intermediate outcomes of exploration and exploitation: The complexity link. Computational and Mathematical Organization Theory, 23(1): 61–93.  DOI: 10.1007/s10588-016-9217-1 (https://rdcu.be/5wsj)

Description. In this paper I suggest an approach to derive probability of organizational success from an intermediate construct, the level of accumulated organizational knowledge. Extent of matches between sub-samples from the org. code knowledge and the environment (or external reality) enable this computation. Thereby, outputs of managerial efforts are better discerned even if environmental perturbations exist. The paper builds on the elaboration by Mosakowski and McKelvey (1997) that the resource-based-view (RBV) of the firm is, in fact, not a tautology, as is discernible when intermediate constructs are separated from final outcome constructs.

What is accomplished? Complexity is introduced as a key construct connecting intermediate outcomes with final outcomes. This enables implementing better accountability of managers.

(5). Chanda SS, Ray S, McKelvey B (2018) The continuum conception of exploration and exploitation: An update to March’s theory. M@n@gement, 21(3): 1050–1079. https://management-aims.com/index.php/mgmt/issue/view/189

Description. In this paper we show that for the continuum conception of exploration and exploitation, March’s (1991) result (Figure 2, p. 77) that more exploration is always desirable reverses if we use a lower stock of collective human capital (CHC) than that assumed in March’s experiments. Our research indicates that a section of extant research is mistaken in assuming that March’s formal model for the continuum conception suggests an inverted U-shaped relation between the extent of exploration and organizational outcome. Instead, the level of CHC determines whether it is rewarding to focus on exploration or exploitation. Thus, the formal model supports managerial intentionality towards exploratory and exploitative innovation through appropriate choice of the level of CHC. We call for a new “balance” discussion, focusing on the determinants of the minimum level of the non-preferred activity from among exploration and exploitation.

What is accomplished? This paper effectively provides the missing half of March’s theory regarding the continuum conception of exploration and exploitation.

(6). Chanda SS, McKelvey B (2018) A Computational Study Explaining Processes underlying Phase Transition.

Available at arXiv: https://arxiv.org/abs/1810.04036

Description. In this article we demonstrate a mechanism for phase transition that has the potential to challenge the dominant Ising model, or inform where the Ising model fails. Here, phase transition is defined as attainment of very widely differing final value on an outcome of interest, on account of small differences in initial conditions. We unearth an elegant mechanism by going deep into results akin to phase transition found in a genetic algorithm model. The mechanism involves initial accumulation of incorrect knowledge (or harmful chemicals / vectors) OR correct knowledge (or beneficial chemicals / vectors) owing to initial difference in concentration, followed by positive feedback loops where (a) a virtuous cycle leads to high accumulation of correct knowledge (or beneficial chemicals / vectors) and (b) a vicious cycle leads to high accumulation of incorrect knowledge (or harmful chemicals / vectors). 

NK Modeling Research (listed in reverse chronological order)

All papers in this section involve computational simulations on Kauffman’s NK Fitness landscape 

(1). Yayavaram S, Chanda SS (2023) Decision making under high complexity: A computational model for the science of muddling through. Computational and Mathematical Organization Theory 29: 300–335. https://doi.org/10.1007/s10588-021-09354-9  (https://rdcu.be/c5uF8 )

Description. In this study, we demonstrate that Lindblom’s decision-making principle of “muddling through” is a very effective approach that organizations can use to cope with high complexity. Using a computational simulation (NK) model, we show that Lindblom’s “muddling through” approach obtains outcomes superior to those obtained from boundedly rational decision-making approaches when complexity is high. Moreover, our results also show that “muddling through” is an appropriate vehicle for bringing in radical organizational change or far-reaching adaptation.

What is accomplished? This paper effectively provides a scientific validation to a decision-making technique, “muddling through” (that helps deal with situations involving high complexity), that was hitherto considered as merely a description of the decision-making approach of the US Congress (or other democratic groupings). Scholars considering commencing NK-model research are likely to benefit greatly from the literature review (and the supplementing appendix) discussing design considerations in the existing NK-model publications in management and organization studies.  

(2). Chanda SS (2021) An Algorithm to Effect Prompt Termination of Myopic Local Search on Kauffman-s NK Landscape.

Available at arXiv: https://arxiv.org/abs/2104.12620

Description. In Kauffman’s NK model, myopic local search involves flipping one randomly-chosen bit of an N-bit decision string in every time step and accepting the new configuration if that has higher fitness. One issue is that, this algorithm consumes the full extent of computational resources allocated—given by the number of alternative configurations inspected—even though search is expected to terminate the moment there are no neighbors having higher fitness. Otherwise, the algorithm must compute the fitness of all N neighbors in every time step, consuming a high amount of resources. In order to get around this problem, I describe an algorithm that allows search to logically terminate relatively early, without having to evaluate fitness of all N neighbors at every time step. I further suggest that when the efficacy of two algorithms need to be compared head to head, imposing a common limit on the number of alternatives evaluated—metering—provides the necessary level field.

(3). Chanda SS, Yayavaram S (2021) Overcoming Complexity Catastrophe: An Algorithm for Beneficial Far-Reaching Adaptation under High Complexity.

Available at arXiv: http://arxiv.org/abs/2105.04311

Description. In his seminal work with NK algorithms, Kauffman noted that fitness outcomes from algorithms navigating an NK landscape show a sharp decline at high complexity arising from pervasive interdependence among problem dimensions. This phenomenon—where complexity effects dominate (Darwinian) adaptation efforts—is called complexity catastrophe. We present an algorithm—incremental change taking turns (ICTT)—that finds distant configurations having fitness superior to that reported in extant research, under high complexity. Thus, complexity catastrophe is not inevitable: a series of incremental changes can lead to excellent outcomes.

(4). Chanda SS, Yayavaram S (2020) An Algorithm to Find Superior Fitness on NK Landscapes under High Complexity: Muddling Through.

Available at arXiv: https://arxiv.org/abs/2006.08333

Description. In this article we show that, under high complexity—given by pervasive interdependence between constituent elements of a decision in an NK landscape—our algorithm obtains fitness superior to that reported in extant research. In our algorithm, we distribute the decision elements comprising a decision into clusters. When a change in value of a decision element is considered, a forward move is explored if the aggregate fitness of the cluster members residing alongside the decision element is higher. The decision configuration with the highest fitness accomplished in the path is selected. Our algorithm obtains superior outcomes by enabling more extensive search, and allowing inspection of more distant decision configurations. We name this algorithm the muddling through algorithm, in memory of Charles Lindblom who spotted the efficacy of the process long before sophisticated computer simulations came into being.

Conceptual Papers (in reverse chronological order)

(1). Chanda SS, Banerjee DN, (Forthcoming) Omission and commission errors underlying AI failures. AI & SOCIETY. https://doi.org/10.1007/s00146-022-01585-x  (https://rdcu.be/c5uFR )

 (https://rdcu.be/ddEzH )

Description. We investigate origins of several cases of failure of Artificial Intelligence (AI) systems employing machine learning and deep learning. We focus on omission and commission errors in (a) the inputs to the AI system, (b) the processing logic, and (c) the outputs from the AI system. Our framework yields a set of 28 factors that can be used for reconstructing the path of AI failures and for determining corrective action. Our research helps identify emerging themes of inquiry necessary for developing more robust AI-ML systems. We are hopeful that our work will help strengthen the use of machine-learning AI by enhancing the rates of true positive and true negative judgements from AI systems, and by lowering the probabilities of false positive and false negative judgements.

(2). Chanda, Sasanka Sekhar (2023) A Constructor Theory-based Approach for Computer Code Model Validation: The Crucial Role of an Effort-directing Feedback Mechanism. Available at SSRN: https://ssrn.com/abstract=4468005 or http://dx.doi.org/10.2139/ssrn.4468005.

Description. I argue that human intelligence is effectively utilized when effort-directing feedback processes are taken recourse to, in any validation task, in order go around the human frailty of giving up prematurely in the absence of actionable feedback. The insight comes from application of a principle from Deutsch’s constructor theory when comparing two alternative approaches for validating prior research containing a computer code model. In one approach, an independent researcher undertakes the set of mental processing tasks necessary to convert natural language specifications in a prior publication’s text into fresh computer code instructions. The numerical or graphical outputs from the fresh computer code are compared with the numerical or graphical outputs found in the original publication. In this case, the original computer code must not be consulted. The effort towards obtaining convergence with the numerical or graphical outputs directs intensive and iterative reconsideration of the publication text, for refinement of the fresh computer code. The second approach comprises going through the original computer code and checking conformance with the specifications in the publication text. This approach is inferior, because inconsistencies between the specifications in the publication text and specifications in the original computer code are quite likely to go unchallenged. This principle elucidated here can be put to use to develop better decision-making configurations in business, government and society.

(3). Banerjee DN, Chanda SS (2020) AI Failures: A Review of Underlying Issues.

Available at arXiv: https://arxiv.org/abs/2008.04073

Description. We find that AI systems fail on account of omission and commission errors in the design of the AI system, as well as upon failure to develop an appropriate interpretation of input information. Moreover, even when there is no significant flaw in the AI software, an AI system may fail because the hardware is incapable of robust performance across environments. Finally an AI system is quite likely to fail in situations where, in effect, it is called upon to deliver moral judgments — a capability AI does not possess. We observe certain trade-offs in measures to mitigate a subset of AI failures and provide some recommendations.

(4). Chatterjee A, Chanda SS, Ray S (2018) Administration of an organization undergoing change: Some limitations of the transaction cost economics approach. International Journal of Organizational Analysis, 26(4): 691–708. DOI: IJOA-07-2017-1202 https://www.emeraldinsight.com/eprint/PFGBTPPAGSZHAJ92RNFD/full

Description. In this paper we build theory highlighting the dark side of transaction cost economics (TCE) theory. Specifically, we discuss the deleterious consequences that ensue when TCE is used to administer organizations undergoing change. The dysfunctions arise owing to (a) TCE’s inappropriate overreliance on managerial foresight (b) TCE’s inability to handle interaction between transactions, given that TCE uses a transaction between dyadic parties as the unit of analysis and (c) TCE’s inability to distinguish between shirking and honest mistakes.

What is accomplished? This paper contributes to the literature that highlights negative impacts of governing organizations by the TCE approach. Specifically, it suggests reasons why large change projects fail and why radical innovation has virtually dried up in multinational firms governed on TCE principles.

(5). Chanda SS, Ray S (2015) Formal theory development by computational simulation modelling: A Tale of two philosophical approaches. Decision, 42(3): 251–267. DOI: 10.1007/s40622-015-0096-y. https://link.springer.com/article/10.1007/s40622-015-0096-y

Description. In this article we distinguish two streams of research for theory development by computational simulation modelling: a critical realist stream that seeks to investigate outcomes by inverting one or more key assumptions in a dominant agent-based model, and a scientific realist stream in the semantic conception tradition that seeks to extend theory or build theory by new constructions in well-known agent-based models, preserving key assumptions.

What is accomplished? The paper lays down a roadmap for researchers wishing to work on theory development by computational simulation modelling (TDCSM). It provides guidance to editors for suitably assessing a TDCSM manuscript.

(6). Chanda SS (2015) CEO cognition in strategy research.

Available at SSRN: http://dx.doi.org/10.2139/ssrn.2586215.

Description. In this article I argue that the mental maps of CEOs get shaped by the experiences they accumulate by meeting various stakeholders as part of their job role, e.g., (a) the TMT (b) Financial Analysts in the Wall Street (c) Regulators (d) Media (e) Board (own firm) (f) Shareholders (g) Debt and Bond holders (h) Employees of the firm (i) Management Consultants engaged by the firm (j) Trade Associations, (k) Board of Director membership in other firms.

Empirical Research

(1). Chanda SS, Ray S (2021) Why Do Strategic Projects Fail?  Available at SSRN: https://ssrn.com/abstract=3836325

Description. There is no dearth of literature inquiring into the reasons for failure of strategic projects—major organization change efforts involving significant, irreversible commitment of resources. However, researchers have failed to go beyond blaming company managers, for example by observing that managers function in silos, protect own turf to the detriment of the project and escalate commitment to failing courses. Yet, in other situations, the same managers obtain excellent organizational outcomes. Thus, there is a need for inquiring into more distal causes of failure embedded in micro-level organizational arrangements.  We study the failure of a strategic project by Burawoy’s extended case method. This approach applies reflexive science to ethnography to extract the general from the particular, facilitating a move from the micro to the macro level. We find that administrative tenets drawing from dominant management orthodoxies—setting direction and settling issues by fiat, rigorously metering accomplishment of organizational members to goals committed to ahead of time, and swift punishment upon deviation from commitments—give rise to dysfunctions that lead to failure of a strategic project. We offer practical advice to managers to help lower the risk of failure of a strategic project. We suggest that, in order to foster the level of coordination necessary in an organization-wide change project, it is necessary that internal and external stakeholders provide reassurance there shall be no sanctions when project goals morph over time as the organization progresses on the execution path through trial-and-error learning. Specifically, relaxation of near-term performance expectations is necessary.

(2). Chanda SS, Ray S (2016) Learning from project failure: Globalization lessons for an MNC. Thunderbird International Business Review, 58(6): 575–585. DOI: 10.1002/tie.21776

Description. In this paper we suggest that as far as configuration and implementation of information technology solutions for companies is concerned, the design skills clearly lie in countries like India, on account of greater familiarity with a wider variety of business processes compared to the Western countries. This is quite the opposite of what transpired when Western countries shifted manufacturing to China, keeping design to themselves.

What is accomplished? The paper highlights that higher variety in service delivery in the emerging markets confers higher service design and configuration skills to emerging market designers, challenging a groupthink that all design must occur in the West.