Focus: Entropy & Homeostasis

Preamble

As defined by thermodynamics entropy is a measure of the energy within a system that cannot be usefully harnessed; cybernetics has took over, making entropy a pillar of information theory.

Figuring Digital Matter (Marcelo Cidade)

Notwithstanding the focus put on viable systems and organizations (as epitomized by the pioneering work of Stafford Beer), cybernetics’ actual imprint on corporate governance has been frustrated by the correspondence assumed between information and energy. But the immersion of enterprises into digital environments brings entropy back in front, along with a paradigmatic shift out of thermodynamics.

domain: Physics vs Economics

The second law of thermodynamics states that entropy within a system is constant, and so is information as defined by cybernetics. But economics laws, if there is such a thing, are to differ: as far as business is concerned information is not to be found in commons but comes from the processing of raw data.

As a matter of fact, thermodynamic stability is meaningless in economics because business information is not a given and uniform quantity but a fluctuating and polymorph one. It ensues that for enterprises in competitive environments measures of entropy set in isolation are pointless: taking a leaf from Lewis Carroll’s Red Queen, the entropy of any given enterprise can only be assessed in relation with its competitors.

Taking a general perspective, entropy is meant to be observed either between the two ends of communication channels, or at the boundaries of complex systems (animals, machines, or organizations). While the former can be seen as a formal coding issue, the latter is of particular relevance for enterprises immersed in digital environments.

Departing from thermodynamics view of entropy measuring a system’s thermal energy unavailable for conversion into mechanical work, economics will consider unexplained information, i.e which cannot be acted upon.

But leaving physics also means forsaking the comfort of its metrics. That’s not a problem for communication entropy, which can be dealt with as a coding issue; but the difficulty appears when counts of digits have to be replaced by measures of knowledge.

complexity: microstates vs macrostates

When there are no clear and reliable metrics to put light on it, entropy becomes a black hole defined by disorder or randomness, the former in reference to the order borne out by classifications, the latter in reference to the predictability borne out by probabilities. Since both approaches deal with the relationship between configurations and information, the issue can be summarized with:

  • A quantum of hypothetical information (dashed line).
  • The complexity of identified items (aka microstates).
  • The complexity of representations (aka macrostates).
  • The quantum of information accounted for, ordered or predicted.
How to assess what is missed ?

As stated by the basic law of thermodynamics, to be of any energetic use the internal heat of a system has to be lower than the external one. Generalized in terms of complexity, it means that the quantum of information accounted for will first increase with the number of configurations considered, reach a maximum, and then decrease until the complexity of configurations (aka internal heat) equates the complexity of the target (aka external heat).

Translated to economics, that scheme must be restated in terms of digital environments and business information.

Entropy: From data to information

The generalization of digital exchanges between enterprises and their environment points to the necessary distinction between data (exchanged flows) and information (managed assets). What happens in-between can be defined along two perspectives:

  • Representation: whether the exchanges are carried out at digital level (data observations of microstates) or in association with symbolic representations (models of macrostates) of managed information.
  • Coupling: whether the processing of data is tied to operations or carried out independently.
Entropy Layers

The issue of entropy can be restated accordingly:

  1. The direct consequence of the digital transformation is to increase osmosis between enterprises and their environment, with data flows bypassing traditional boundaries and feeding directly operational processes (bottom right).
  2. As a result, data analytics can be integrated with business processes, improving operational decision-making. In terms of entropy the outcome would be a more effective use of information models (bottom left).
  3. Next, information models (aka symbolic representations) can themselves be changed as to improve their effectiveness, i.e reduce entropy at process as well as architecture level (top left).
  4. Last but not least, deep learning could be applied to digitalized contexts (aka microstates) in order to derive new symbolic representations (aka macrostates) (top right).

On that basis entropy policies should be set at operational and strategic levels, the former dealing with processes, the latter with architectures.

Homeostasis: from Data to knowledge

As far as enterprises are concerned, homeostasis means the adaptation of organizations and systems to changes in competitive environments. To that end, policies set along the entropy paradigm could draw on improving the perception of microstates (data level) as well as the representation of macrostates (information level).

At data level, the perception of changes depends on the osmosis between systems and environments. To that end data mining is to be used to hone the sampling of populations and refine microstates.

At symbolic level (representations), the economics perspective differs from physics on two critical aspects:

  • Contrary to heat, economic information is not a constant, which means that macrostates are to be continuously updated.
  • Moreover, the economic playground is not homogenous but combines physical (e.g demographics or climate), socio-economic (e.g income or education) and symbolic (e.g politics or culture) microstates, which means that macrostates are to be differentiated.
Improving entropy

With or without differentiated macrostates, policies are to rely on two categories of models:

  • Extensional ones target environments, dealing with the analysis of business context, operations, and changes (descriptive models), as well as what should be expected (predictive models).
  • Intensional ones deal with enterprise architectures, current or planned, at processes as well as architecture level (prescriptive models).

As figured above, representations and policies are to be combined, hence the need to anchor them to enterprise architectures, as can be illustrated with the Pagoda blueprint:

Architectures & Homeostasis (crosses and numbers refer to the table above)
  • Populations of instances (microstates) are obtained from digital environments as well as from processes execution (1). Data analytics are used to improve descriptive and predictive models (operational macrostates), and consequently the osmosis between processes and environment (2).
  • Architecture improvements can be achieved through changes in prescriptive models (structural macrostates). Compared to strategic planning carried out top-down from business models and requirements, homeostasis also relies on bottom-up forces which can combine with business models and coalesce into emerging architectures (3).
  • Business and organization models are arguably a primary factor in sorting out data, respectively from environments and systems; and being by nature essentially symbolic, they are more pliable than systems models (0). Homeostasis could therefore bypass architectural macrostates, with models of business and organization emerging directly from digital environments through data mining and deep learning (4).

Such alignment of enterprise architectures and symbolic representations is to greatly enhance the transparency and traceability of digital transformations by organising changes and policies with regard to their nature (physical, functional, organizational, or business) and decision-making horizon (operational or strategic).

FURTHER READING

One thought on “Focus: Entropy & Homeostasis”

Leave a Reply

%d bloggers like this: