Tuesday, June 08, 2010

Embedded versus Embodied Decision Support

Within the realm of decision support methodologies, two very different paradigms are vying for the attention of enterprise. The first is what I call the embedded approach to decision support. The embedded approach to decision support emphasizes scientific logic and rigor, and is grounded firmly in the traditional disciplines of operations research, systems analysis, decision analysis, and risk analysis[1].

Embedded Decision Support Methodologies

The second still emerging approach is what I call the embodied approach to decision support. The embodied approach to decision support traces its roots to the information technology movement and enjoys critical acclaim for its potential for automated performance monitoring, business intelligence, and business analytics.

Embodied Decision Support Methodologies

Note that the content-validity of the embedded approach is widely accepted amongst professional researchers and analysts as a body of knowledge. The literature underlying the embedded approach is vast and rich in empirical evidence supporting the validity and reliability of its methods. This extant literature regarding the embedded approach is synonymous with the disciplines of operations research, systems analysis, decision analysis, and risk analysis.

The content-validity of the still emerging embodied approach remains in question. Researchers and analysts are still debating many of the terms used in the embodied approach, and a broad consensus regarding what exactly business intelligence and business analytics entail is not yet evident. The existing literature supporting the effectiveness of embodied methods is mostly descriptive with scant empirical evidence to support its validity and reliability as a proven decision support methodology.

The significance of the differentiation between embedded and embodied methods lies in the warranties that each provide the decision maker. An impressive quality of the embedded approach is that all the terms and concepts used are clearly defined and widely accepted by professional researchers and analysts thus enabling users to articulate universally their findings and recommendations.

In contrast, the lack of consensus regarding the validity and reliability of embodied methodologies limits the utility of what we know to be business intelligence and business analytics. Indeed, the methodological frameworks for business intelligence and business analytics are still emerging in the form of dashboard reporting systems and other untested visualization methods that some researchers argue can lead to cognitive distortions of the evidence uncovered by such methods.

Future consilience between the practitioners of embedded and embodied methods is far from complete or even certain. More empirical evidence will be needed before the embodied approach can be fully converged or enjoined within the deeper conceptual foundations of embedded methodologies. In the mean time, decision makers are advised to take precautions to ensure that embedded methodologies take the lead in verifying and confirming the findings and recommendations of embodied technologies.

[1] Note that risk analysis is not to be confused with risk management, which is a different function and discipline all together.

2 comments:

Scott Mongeau said...

Greatly appreciated: touches on struggles I'm having in defining the emergence of a fluid set of overlapping disciplines.

One potential overlap at the end of each curve is a point where Management must resolve formal Decision Management conflicts and paradoxes utilizing the emerging field of Multiagent systems; and capping that thematically, perhaps simply leadership in the age of Cyborg Organizations.

At a Risk Management conference recently, proponents of the 'embedded' school were vying for creation of a PMI-style professional association. However, one central tenet was the notion that 'Decision Professionals' should not be held responsible for the outcome of their decisions as long as correct Risk Analysis principles are applied.

What is left at the bleeding edge where Risk and Business Analytics absolve themselves, even in the aggregate, of any positive or negative outcomes? In the case of Risk Analytics, practitioners wriggle from responsibility by ascribing religious power to a set of theoretical principles. In the case of Business Analytics, the same can be said, albeit one can additionally point to 'the machine' and add that not only theory, but encoded theory absolves advice from outcome. Both, to me, seem to be an attempt by crypto-bureaucrats to insulate themselves professionally for the repercussions of their actions (where, in this frame, giving advice is an action with potential formal consequences).

What is missing then? It appears in both cases decisions are thrown to Management, who are expected to sort through the advice and take a stake in the risk and to bear the consequences.

The growing problem implied by the projection of responsibility by the Embedded and Embodied arcs is that Management (as a formal discipline itself) increasingly is beset with multiple such arcs bestowing varying advice (theoretically-based) from multiple decision professionals and systems in the prototypical transnational.

I'm wondering if both arcs meet at the emerging discipline of Multiagent Systems? It is here that multiple, often conflicting, sets of theoretical advice and systems converge on a 'game board', and where Managers are expected to choose the best option.

As Behavioral Research and Economics increasingly makes clear, professional Managers are often ill-suited to resolving the inherent paradoxical advice from such multiple systems. Multiagent systems offers a formal methodology for Management to second-guess the pitfalls of Behavioral bias by integrating Game Theory and Decision Management to consolidate advice from conflicting participants and systems.

The challenge then seems to be: how to train a new generation of Managers and business leaders suitably to understand the context and challenge of Multiagent systems? One roadblock is that the extroverted and emotional personality profiles that statistically are attracted to (and advance within) the Management profession are often adverse to the highly analytical processes that must be applied to formally resolving Multiagent problems.

Perhaps the forward problem is exploring the challenge of leadership in the coming age of Cyborg Organizations, where large institutions are increasingly sloppily bound together by unruly and conflicting sets of formal advice from disparate perspectives (be they theoretical, system encoded, emotional, etc.)? Without wishing to sound hyperbolic, is this problem indeed the capstone of a large and lingering problem (particularly over the last century as populations expand) of the human species: how can we ensure that leaders make optimized decisions with consequences for large groups when the scale and breadth of the problem-sets are increasingly alien to the capacity of individuals and small groups to encompass formally via traditional behavioral-based individual judgment perspectives?

Dr William J McKibbin said...

Hi Scott, thanks much for your incisive and very interesting perspective and comments -- you got me thinking...

Post a Comment