Our work on foundations for the Virtual Enterprise assumed a certain vision of an advanced enterprise. It was advanced in at least the sense that the required support is unachievable by current engineering methods. Work towards new foundations to enable novel engineering methods is to my mind the only bona fide research we can do for commerce. And this presumes advances in the science of enterprise infrastructure. In Europe, for historical reasons these are considered problems of interoperability.
The problem is that for organizational reasons and an accident of market forces, there is no current science base for enterprise infrastructure. In the US, research on this has been terminated as a result of political ideology: a belief that government should not ‘support’ the market.
Work in the field generally is performed by engineers and the academics who support them. EU program management in their advanced enterprise research program (Future Internet Enterprise Systems, FInES) recognizes this dearth of scientific foundations and has sponsored work toward a relevant science base. Because of the contracting process and various legacies, the work is being performed within the existing community, which is to say by engineers and practitioners of the current state of the art. These are intelligent and dedicated thinkers, but they are the old guard: engineers and not scientists. The result as it is appearing is unlikely to lead neither to a true science base nor a radical advance.
I served on the initial advisory committee and now on the science experts panel. This note is a summary of my recommendations.
A high level view has the science behind any practice as the laws that govern that practice within a coherent framework. When applied, the laws must be capable of predicting behavior when previously unknown conditions are encountered. These laws must be expressible in formal terms (commonly understood as mathematics) and the framework must be coherentlycausal (commonly understood to be logical).
This small note mentions four contributing sciences. Then we provide three research domains and some suggested topics within each. Finally a perspective on layered science prioritizes these.
Enterprise Infrastructure Science
In our experience, science is simply a matter of finding the right abstractions to be drawn from the field of interest and additionally finding the right second order abstractions to properly manipulate them. The first is usually characterized as ‘models’ and the second ‘laws.’ For better or worse, enterprise infrastructure already has a robust notion of models; in fact a good characterization of the practical problem is that there are too many models using too many orthogonal abstractions and ontological frameworks.
Regarding existing models, it is clear that we have to build on the models that currently exist, enhancing them in ways that are supported by the new science base. A rough analogy is the ball-and-stick model of early chemistry; wildly misleading, it is still with us as a way to introduce novices and as a basis for one visualization technique.
Most models within the information infrastructure domain have an explicit methodology, but none currently has a robust formal basis in the way they are applied. Such models include process, product, cost, RIO, resource, logistic, market, economic, collaboration models and so forth. When these are reduced to quantitative abstractions, the formal basis becomes more rigorous because it collapses to simple arithmetic and (occasionally complex) probability. But in that case, the abstractions by definition lose their ontological basis, and utility for our science base vanishes.
So the general challenge is to embrace and extend existing models, rationalize them into a formal scientific framework, federate them ontologically and devise useful user interface conventions.
We do not have an enterprise information infrastructure science, but there is burgeoning science in four relevant areas and these need to be incorporated, and possibly subsumed. A post on enterprise integration is here.
One is the science of management. Business schools (now always associated with universities) developed around understanding commerce and administration rather than the issues of managing an enterprise. Only recently, within the last twenty years, has the idea of managing the enterprise been a concern. Unfortunately, in most cases, the tools from commerce were applied, with the result that enterprise management as taught in those schools is accounting-heavy. Examples are ‘activity based costing’ and ‘economic value added™’ systems.
MIT and General Electric were leaders in developing an actual science of management. Butmost business techniques are taught using the case based method and the qualities of ‘leadership.’ The case method is simple: you study a real life business example and are taught to appropriately emulate or avoid the decisions highlighted therein. The ‘leadershipmodel’ is particularly vacuous, driven by slogans and platitudes. Neither lend themselves to the scientific method.
Nonetheless, there are some emerging trends in management that can be called scientific. These must be incorporated in the new enterprise information infrastructure science base in three levels.
There is the management of the enterprise of course, and this includes the management of the managers and their metrics. There is the management of the information infrastructure, which if it is properly science-based, conceptually rides on top of the enterprise proper. And there is the level of managing the science base as well.
As with all four of the contributing sciences, the parts that truly are scientific and applicable should be subsumed. We deal with that in our summary of topics, below.
Computer (Information) Science
A second emerging science is the science of information systems. This also is properly taught in only a few universities as most teach the tradecraft only. Computer science has a long tradition — and is the level of science we are interested in. There is a rigorous foundation from Von Neumann and a significant cadre of mathematical logicians. But just as management science has been often reduced to numbers and linear algebra, computer science has been largely reduced to the methods of Aristotelian logic and its elaborations.
However, there is a solid mathematical basis of information transformation in category theory and the Curry–Howard correspondence (of mathematical and code elements). There are similar relevant mathematical principles in set, group and graph theories that can support a science base.
For this reason, information science should not only be subsumed (like the others), but also provide the mathematical kinematics. Severalgroups may inform this work.
For as long as theories have been applied to human behavior, the so-called ‘soft sciences’ have attempted to apply the same rigor as physics. We have plenty of observations, but a true science would build theories to both explain and predict. Mostly, social science is at the level of extracting and categorizing statistical patterns from the observations. But there do seem to be consistent patterns that we need to fold into the larger science of enterprises.
Said another way, societies and individual humans bracket the enterprise in scale, and whatever we have in terms of science must be preserved in the larger scientific perspective of the enterprise science base.
Much the same can be said of economics, and in fact management, economic and social sciences can be considered three facets of the same domain, with computer science constraining and defining the relevant science of logic.
Other universities have ‘business schools’ that grant accounting-inspired degrees (Masters of Business Administration) but MIT alone has a ‘School of Management Science,’ granting a Master of Science in Management degree.
General Electric was the first large corporation to think not of profit-by-any-means as a core driver of decisions, but core competencies as something to master. Profits are the result, not the metric.
For example, the aircraft engine business requires a baffling mix of partners who are sometimes competitors or customers. GEAE looks at aircraft as engines with seats attached. All else is support, businesses that they don’t need to be in and would rather not.
Within the engine itself, the core competence is the hot section. GE is willing to partner for anything else but never this.
The concept of leadership of this type was developed by the military, knowing the principle is largely bogus. It was originally a social engineering device to ascribe a quality that solders could use in discarding qualms about following top-down orders.
It is easily adopted by business schools in large part because the qualities that make a leader are supposed to make up for the lack of logical scientific principles. And it lends itself to the attractive notion that managers have some elite qualities.
Three promising groups are of interest. They are:
Foundations of Information Science
Information science itself is undergoing a serious re-examination, based on some well-understood limits of current scientific frameworks, notably in quantum physics, biology and economics. An interesting group is the Foundations of Information Science group which hosts a long running email list and occasional conferences. The binding presumption behind this group is the idea that information plays a fundamental role in the universe, across disciplines. It is merely an accident of history that we focused on objects and then had to presume forces and fields.
Information, if considered as a causal first class citizen in a process oriented framework, may provide a basis for complementary theoretical frameworks for a science based on organization. My own interest is the enterprise, defined in layers from society, through enterprises, groups, selves, microbiological systems, and chemistry/physics (‘chemistry‘ here is meant in the legacy sense).
As with most such groups, it is held together by comity and the use of legacy tools. Many of the participants are scholars of past work, and one cannot expect the same conceptual tools to produce radical new insights. But they have the agenda right, they have a long history and some very clever thinkers.
A completely independent group explores new geometric logics, a larger class of logics that contains the cases of quantum and Aristotelian logics. These are based on just the sort of advances in category theory that edge on computer science, and can easily be considered the next generation mathematics for working with information. The group self-referentially calls itself Quantum Interaction and has yearly conferences. Applications are many but deliberately exclude physics.
Most participants use regular old quantum logic, but there is also some very advanced work on new linear logics which we would call geometric ‘causal logics.’
This originated as the founding concept of the Stanford Center for the Study of Language and Information. The theory has since been extended and applied to general phenomenon. The basic idea is that ordinary logic — what we have been calling Aristotelian logic — has a number of limits when applied to real world reasoning by humans. This can best be handled by one ordinary logic for how we logically reason about facts and a second reasoning system for how we ‘reason’ about situations. Because this is a reasoning system, and because the two reasoning systems must be integrated, we can this a two-sorted logic.
One, the ordinary one, is based in well understood set theory while the other uses category theory — in fact, possibly some of the very same category theoretic notions mentioned above.
In the DARPA work and following, we found it useful to divide the challenges into three groups, representing:
- the basic elements of the enterprise,
- the way you reason about the enterprise (and the enterprise ‘reasons’ about itself), and
- the way such reasoning supports the enterprise as a whole system.
Basic Elements of the System
The first of these was mentioned above as the problem of modeling. We don’t model the right things; what we model cannot be federated; and none of it has a rigorous science base. So the first group of problems has to do with representing what the enterprise does. We already have product, process, resource, financial and various kinds of accounting models, but none of those deal with what the components of the enterprise actually are supposed to do: incrementally add value in context.
The research challenges are to develop a set of abstractions that adequately models the value contributed to the enterprise as a whole system in an effective number of infinite possibilities. This value would be associated with any element or aggregation in the system, for example the common resources and processes, as well as more implicit qualities like urges and intents.
Value abstractions would not be scalar but functions that would modify themselves based on the context of others. In the context of the logic and interaction below, they will have agency and allow some degree of self-organization. This implies that they will not be set-theoretic in the strict sense.
Such abstractions will subsume product, process and collaboration models and of course enrich existing quantitative models.
The Logic of the System
The ‘logic’ of the system has to explain and predict the behavior of the system in the real world and how that world responds to it. Clearly, ordinary logic and any parametric frameworks are inadequate. Some reasoning over unknowns is required. Some way of capturing emotions and intuitions is necessary, like we sometimes associate with brand, lifestyle and delight. Some apparently quantum logical effects we see in the market and among humans and tribes must be accommodated.
Since the enterprise in its many dimensions encompasses all of reality, it must have the ultimate-spanning reasoning machinery. Since the enterprise is something that we both create/manage and that we are immersed in, the logics should be deeply introspective but also manage the blind spots that we know about (and of course those we don’t).
This is to say that the research into new logical foundations needs to provide what has hitherto eluded all of soft science. Possibly some extension of Markov-Bayes will show promise, some development from non-monotonic reasoning, or some maturation from situation theory.
The Interaction Dynamics
This component deals with behavior: how the logic of the system operates with the elements to explain all known behaviors of the enterprise at all levels. Some levels are molecular, as with new materials or pharmaceuticals. Some are Newtonian, such as logistics and schedule dependencies. Some are quantitative because that is what the financiers (currently) understand. And some operate at the human and societal levels in the aforementioned economic, managerial and social domains.
All of the promise from complexity theory needs to come from this, despite the practical collapse of that theory. Self-organizing and agent system dynamics of various kinds needs to be explained.
As if this were not sufficiently impossible, the dynamics need to predict essential value results in unknown or previously unencountered situations. (This is a requirement of science.)
Because we are talking about information infrastructure, we need to be able to directly map the statements of dynamics to functions in code. Because the system needs to be introspective and self-modifying, we can have no essential distinction between code and data (which is the same as saying that enterprise design and management is incorporated in the enterprise).
(A brief presentation on these topics was recorded at the FInES workshop on Samos, Greece in July 2011. Email me for the hard-to-navigate-to link to the video.)
All else is engineering. The challenges extend to the grandest of grand challenges in science. A mitigating factor is that the tiniest progress can have profound effects, because we are so very ignorant of enterprise dynamics.
An epistemological proposal will follow, based on the principles themselves as an emergent system.