Knowledge-driven Decision-making (Part 1)

Decision Tree & Learning Path (Fabien Merelle)

Decision-making is supposed to be informed, and for enterprises that can be best achieved through ontologies.

The primary aim of ontologies is to weave together thesauruses, for the meaning of words (business domains), and models, for the representation of environments, organization, and systems. But ontologies can also deal with the epistemic status of representations: actual, past, planned, expected, hypothetical, virtual, fictional, etc. To take full advantage of these capabilities ontologies must form the nexus of business intelligence and decision-making, and that can be done by gearing knowledge- and decision-making processes.

This first part introduces the cogs between Observation, Orientation, Decision, and Action (OODA) on the one hand, Data, Information, and Knowledge on the other hand. The second part will detail the learning roundabout.

see also: Knowledge-driven Decision-making (Part 2)

Informed Decision-making

Assuming a knowledge management framework, the objective is to pair (as in DNA strands) the Observation/Orientation/Decision/Action (OODA) steps of decision-making processes with their Data/Information/Knowledge counterparts for content.

OODA Steps

The Observation/Orientation/Decision/Action (OODA) loop provides a basic blueprint for decision-making processes:

  • Observation: data can be obtained from digital or business environments. It can be analysed and turned into knowledge (business intelligence) or matched with models of managed information.
  • Orientation: reasoning (information) is applied to observations (data) and judgment (knowledge).
  • Decision: judgment (knowledge) is supported by observations (data) and causal chains (information).
  • Action: decisions are carried out through a mix of operations (data) and organization (knowledge).
Informed OODA

Decision-making sequences can then be characterized in terms of knowledge management and timescales.

Overview of OODA representation

Augmented OODA

Decision-making is often confused with problem-solving, namely how to pick a solution given a set of resources, typically people, information, financing, materials. That paradigm ignores the temporal dimension of enterprises’ decision-making which are made of interdependent commitments meant to be carried out across shifting backgrounds and overlapping timescales. That temporality makes room for changes in decision-making contexts, typically:

  • Uncertainty (observations): facts are not gold nuggets ready to be picked from river beds; their meanings are mined from data and applications as determined by segmentation (models). Confidence, and more generally the quality of data, can thus be enhanced during decision-making processes.
  • Causalities (orientation): competing in digital environments means that root causes and rationales, usually set upfront for problem-solving, can now be reassessed on a continuous basis, with causal chains improved through the integration of data analytics and information models.
  • Risks (decisions): since business competition is by nature a time-dependent, nonzero-sum game, risks can be reassessed until the “last responsible moment,” when delaying a commitment would reduce the range of options or the expected returns. Such reassessment (knowledge) can take into account reduced uncertainties (data) and improved causal chains (information).
  • Experience (action): decisions taken at the enterprise level usually involve sets of intertwined commitments and deeds whose efficiency is determined by the alignment of operations (data) with organization (knowledge). Set in digital environment, experience can benefit from direct feedback combining knowledge and observations.
Augmented OODA

As to make the best of these changes, basic sequences can be factored out according to their impact on control variables:

  • Commitments: decisions can be made with suspended execution when delays can help with the uncertainty of observations, the reliability of causal chains, and consequently the adequacy of risks assessment (a).
  • Routine: decisions can be carried out directly based on observed data, without a reassessment of already secured information and causal chains (b).
  • Automation: actions can be carried out automatically (bypassing organizational considerations) based on reliable observations and confirmed causal chains (c).
  • Management and support: decisions about the use of resources could be made and actions carried out without observable impact on environments (d).

More generally, these patterns can also be used to characterize decision-making horizons:

  • Operational: when observed data can be directly mapped to information and put to use as knowledge, thus allowing for routine decision-making (a)
  • Tactical: when the execution of decisions can be improved by additional data and better traceability (b)
  • Strategic: when commitments must be made without reliable causal chains for the time frame considered; regular costs/benefits assessments must thus be replaced by risk-management schemes covering for ill-fated turns of events (c)
  • Management and support: for actions meant to improve the capabilities of enterprises’ organization and systems.
Horizon Knowl’Edges: Operational (a), Tactical (b), Strategic (c)

That brings back the temporality issue: how to manage decision-making horizons in digital environments characterized by continuous changes in markets, competition, and regulations. The answer is to weave together decision-making and knowledge management.

Decision-making Roundabout

The digital revolution comes with both constraints and opportunities for enterprises decision-making. On the one hand decisions with lasting commitments must be taken in shifting environments with blurred horizons; on the other hand the weaving of software components into business processes greatly enhance their adaptability, especially when AI and ML technologies provide learning capabilities.

Assuming that decision-making processes are supported by knowledge management (KM), they could benefit from learning capabilities provided by Machine learning and Knowledge graphs technologies, the former to help with data, the latter as a conceptual glue between environments and enterprises’ information models.

Enterprise Architecture & Knowledge Management

More specifically, the pairing of OODA steps with KM enable dynamic improvements of observations, reasoning, judgments, and experience:

  • Observation: data is not given but obtained through designed apparatuses; the relevance and quality of observations (cf. data mining) can thus be improved (a) during decision-making processes; improvements can also be achieved through feedback (cf. process mining) from modus operandi (e)
  • Orientation and decision: models and the reliability of business causal chains can be improved when actual outcomes can be compared to expected ones (b); the same approach can also be used for the assessment of policies (c)
  • Action and observation: given a decision, modus operandi can be improved through the analysis of actual and expected resources and execution (d)
Adding learning capabilities to decision-making

With regard to enterprise governance the pairing of OODA steps with topics (mined data, models, policies, modus operandi) sets the frame of individual and collective learning fusing organization and systems. More generally the pairing decision-making processes with the processing of data into information opens the door of a cybernetic approach to enterprise architecture.

Viable Systems & Learning Organizations

Taking a leaf from Stafford Beer, the primary objective of digital transformation is to enable the continuous and timely adaptation of enterprises to their environment. When these environments are shifting as well as competitive, adaptation must be carried out at two levels: systems and organization.

The Matter of Time

Compared to problem solving, decision-making at the enterprise level is generally set in time frames: on the one hand such decisions are often based on data whose reliability vary with time; on the other hand they involve a number of commitments to be carried out along time.

Temporal databases deal with that issue by tying time-sensitive entries to axes of time set by defining events (aka epochs), typically for observation, resources shelf-life, and commitments. But dealing with time as a linear succession of events may be enough if temporality bears only on data, not if it bears also on causal chains and judgments, and consequently on traceability and accountability. In that case what matters is the combined history of contributing elements: meta-data, causal chains, risks. Such history could then be turned into a collective memory, experience, and learning.

Augmented OODA & Cybernetics

As a playback of John Boyd’s experience as fighter pilot and military strategist, the renewed relevance of his OODA brainchild comes from the correspondence between the seamless integration of IT systems with business processes, and fighter jets’ command and control processes. A correspondence now boosted by AI and ML technologies. On that account the pairing (as with DNA strands) of OODA steps with data and information processing can be a game changer as it enables a conceptual as well as practical integration of EA in cybernetics.

With regard to environments, the interoperability of Machine learning and Knowledge graphs enables a seamless integration of action, observation, and orientation steps, taking full advantage of the digital osmosis between operations and environments.

A Cybernetic View

With regard to enterprise ontologies enable interoperability of Knowledge graphs with thesauruses and database management systems, paving the way to enterprise’s homeostasis and knowledgeable organizations.

Organization & Knowledge

The ontological integration of all symbolic resources and their interoperability through OODA decision-making processes, set the ground for comprehensive learning capabilities across systems and organization:

  • Collective, because judgments should be backed by transparent and traceable knowledge and reasoning, independently of individual beliefs or opinions
  • Individual, because when judgments translate into decisions, even ones taken collectively, their traceability also implies personal accountability

That pivot from learning (individual and collective knowledge) to decision- making (individual accountability) is arguably a critical issue for organizations. It can be analyzed with regard to the problem at hand, or from the broader perspective of organizational behavior.

Pairing decision- and knowledge-making processes begets a quadrant of learning capabilities crossing the kind of agent (people or systems) with the kind of knowledge (implicit or explicit):

Learning Capabilities

Transitions from implicit to explicit knowledge can thus be achieved at individual and collective levels:

  • Between people and organizations, it’s typically done through a mix of experience and collaboration (a)
  • Between systems and representations, it’s the nuts and bolts of Machine learning and Knowledge graphs technologies (b)
  • Between people and systems, learning relies on the experience feedback achieved through the integration of ML into the OODA loop (c)
  • Between organization and systems, learning relies on the functional distinction between judgment, to be carried out at the organizational level, and observation and reasoning, supported by systems (d)

Such integration of systems and organization will ensure:

  • Knowledge based collaboration
  • Accountability (organization) and traceability (systems) of decision-making and reasoning processes
  • Incremental and smooth learning curves
  • Learning driven by people experience and feedback from systems

That would transform enterprises into viable organizations with the “ … ability to learn and translate that learning into action.”  (Jack Welch)


%d bloggers like this: