Bo Strangert (RD15)

Control-theoretical reasoning as a means to structure

complex realities: A case of healthcare planning 

A recent paper in Swedish on this website argued for the use of control-theoretical reasoning in project planning (M6). The present notes summarize some of the arguments. An underlying motive to the arguments is that behavioral approaches nowadays often seem to underestimate the structural complexity of social practices. However, it does not mean that there is a state of affairs for grand substantial psychological theories, nor is inductive reasoning based on statistical design and data collection a realistic alternative in many cases.

A suggested contrasting approach involves a series of integrated actions of concept development and experimentation, originating from some conceptual structure that allows modeling of essential qualities of complexity (e.g., RD13, RD14).

In short, some rudimentary model is chosen that can be gradually developed and specified according to the unique qualities of individual and situational cases at issue. A suitable model should be sufficiently general to include a diversity of individual elements but also capable to structure the elements to reveal possible interconnections. In this respect, qualitative modeling must consider the dynamics and uncertainty of complex social practices.

Such an approach can have two obvious advantages. One is that you may resist the strong and inhibitory influences of prevailing points of view and culture in the field. The other is that a suitably chosen theoretical point of departure can facilitate the empirical examination of facts by providing inventive steering hypotheses for data collection.

Background to the illustrative case study

This early case study, reported in M6, was chosen because it started with an abstract model-theoretic approach by a young research team who was quite inexperienced with the actual project area: ”Improvement of long-term healthcare of patients”. The project directives emphasized the need to apply a ”holistic view” of patients – to consider not only the physical disabilities but also the psychological and social aspects of long-term healthcare, including the individual patient’s rights and self-determination.

The motivation of the team was a reaction against the overwhelming pressure from the practicians to accept the conventional wisdom and expertise in the field. As an alternative, the team wanted to get a fresh look at the practices by applying theoretical and methodical tools of the advancing cognitive and organization sciences.

The basic research idea was that the planning of healthcare had to manage the diversity of disturbances within and between individual patients. It would require flexible healthcare plans for diagnosis and treatment of physical, psychological, and social qualities of life. A patient’s own reduced capability to regulate important life functions should be dynamically supported through regulatory compensation from the healthcare system.

Preliminary modeling

The core process of healthcare was supposed to be the regulation of the patient-nursing staff interactions, including physical as well as informational and psychosocial activities. The diagnostic situation is characterized by uncertainty about both symptoms and causes of a patient’s incapability to regulate functional disturbances.

We chose a simple ”abstract machine model” from the cybernetic literature to make a rudimentary representation of a patient’s regulatory system (Ashby, 1956). Figure 1 shows how some possible data sources are related to the basic regulatory mechanisms (M6).

Figure 1. Possible sources of data in the patient model

(adapted from Ashby, 1956).

In Figure 1, T represents a given system of transformations to cope with different states for an organism. The outcomes should be within a set of ”good” essential variables, E. However, a set of disturbances D (physical, psychological or social) can threaten positive effects on E, which must be counteracted by a regulatory system, R. The regulation of T can either be proactive (through input from D directly to R), or ”error-controlled” through feedback from negative effects of E to R. A third possibility is that a failing transformation in T may have a direct compensatory coupling to R.

Ashby also defined a control concept of higher-order, C, which in this connection corresponds to a patient’s conscious expressions of will.

Figure 1 shows possible data sources that in principle can be used to infer a patient’s problems with regulation and need for compensatory healthcare actions.

Besides establishing a preliminary structure of data for the design of healthcare planning, a conceptualization in terms of information theory also introduces a way of measuring information. The measure represents the variety (entropy, H) of an information source. The capacity of regulation is the difference between the variety of the disturbance source D and the variety of the regulation capacity R. Ashby’s well-known law of requisite variety expresses this exactly in the equation

H(E) ≥ H(D) + H_D(R) - H(R),

where H_D(R) is a conditional statistical uncertainty about the relation between R and D. Thus, H(E)’s minimum is H(D)-H(R) in the special case H_D(R)=0 (i.e. when R is a strict deterministic function of D).

On handling complexity. The patient model seems true to be very simplistic. However, the purpose of the model was not to describe patients in terms of specific content variables. It was rather an attempt to look at the regulatory mechanisms freshly. Note that a patient’s problem would initially be categorized as indications of some kind of a disturbance D_i to be diagnosed. The indications were based on a patient asking for help and through observations by healthcare personnel. The preliminary description should be open and preferably include various suggestions and hints (e.g. about possible medical, psychic, social, and environmental aspects).

In contrast, the T-system denotes a professional and multi-dimensional account of the states of a patient in the situation, with a focus on given human mechanisms for handling oneself in the situation. The diagnostic knowledge should cumulate dynamically during the period of treatment. The effects on the essential variables E can be categorized through both professional and more unsophisticated indicators. The same means can be used for characterizing the focal regulatory mechanisms R.

Although simple, the chosen model approach could accommodate many different types of indications and elucidate the regulatory structures as well.

The design of healthcare planning for long-term patients

The patient model was a central part of a theoretical meta-perspective on healthcare planning. Thus, professional support was primarily seen as reinforcing and compensating a patient’s own needs of regulation. Control-theoretically, this requires the design of a control structure for diagnosing and compensating the patient’s shortcomings of regulation. But an essential restriction was that the control structure should be congruent with the patient’s stated or implied own will.

This explicit intention demands continuing efforts to fulfill in individual cases.  But it is also an inherent source of conflict between different or vague interests. Therefore, the question about how to design the control structure has no simple answer but has to be confronted recurrently at work. That conclusion led to some consequences for the design of procedures for healthcare planning.

Is it possible to amplify the regulation by combining information from a patient and a nursing staff? Ashby mentioned the possibility of generalizing the simple regulation (as in Figure 1) to a large system but never presented a formal solution. This was made by Aulin-Ahamavaara (1979), who showed how entropy can be decreased by a hierarchy of regulation and control. He states the optimal regulatory capability of such a system with n regulators as

H(R) = H(R1) + H(R2) + … H(Rn),

The system’s regulatory effect is reduced by the uncertainty of regulation,

H_D(R) = H_∑D(R) - H(C),

where H_∑D(R) is the total uncertainties of regulation of the regulators, and H(C) is the total reduction in uncertainties by various control units (C) in the system. The control units function as coordinating actors on higher levels which reduces disorder by an amount of its own control entropy, for instance, H(C_jk) for a control unit j of R_j on hierarchical level k.

Then, Aulin-Ahmavaara applies his ”Law of Requisite Hierarchy” to the special case that the optimal regularity capability H(R) and the uncertainty of the regulators H_D(R) are both constant. The equation shows that the only way of improving the effective regulatory capability is to increase the hierarchy of control, thanks to H(C),

H_eff(R) = H(R) - H_D(R) = H(R) - H_∑D(R) + H(C).

Aulin-Ahmavaara concludes that there will be a tendency to increase hierarchy with constant regulators for better survival of the system.

Yet, an increasing hierarchy of a control structure may not be without problems. For example, when professional diagnostic results cumulate, they may be at variance with a patient’s judgment. This can bias the professional decision-making to the patient’s experienced disadvantage because the healthcare personnel regard their professional competence as superior and superseding the patient’s view in most respects.

One possible solution is to establish a hierarchy of decision-making that is attentive to the patient’s view at the lowest level of care interaction. If necessary, it should be counterbalanced according to best practices of science and professional experience. Such a practice may work for physical matters, though is certainly equally appropriate for ”soft” personal and social states, where the patient regulator (C) ought to dominate.

Hence, an interesting practical and theoretical question is about the consequences for healthcare planning of ”cooperation” between patients and professional staff. Is it possible to improve the regulators on the lowest levels of healthcare, and which would the theoretical consequences be of that? Then the optimal regulatory capability H(R) increases, and their uncertainty H_∑D(R) diminishes.  Consequently, the need for control entropy and hierarchy decreases. Practically, it means that one important task could be to improve healthcare planning for active patient-staff interactions. That could advance the psychosocial climate for the patient in consonance with the demands on ”holistic patient care”.

The preceding remarks outline some of the theoretical reasoning before our encounter with the empirical reality of healthcare planning. The question is now: Was it was worth its effort?

An initial control-theoretic approach as a preparation

for empirical exploration

There are some arguments worth considering for a preparatory control-theoretic approach. One psychological advantage is the mental preparation that comes from the initial formal reasoning about a task; this experience guided us at our first confrontation with the reality of practical healthcare.

There is also a strong rational argument about the advantage of highlighting the characteristics of complexity by choice of suitable models. A deliberate choice advances the formulation of testable hypotheses of complex system features. A common counter-argument is that a formal theoretical approach biases the empirical exploration and will be counterproductive to creative inductive research (e.g. using ”Grounded Theory”). It can be argued against the inductive approach that it is very vulnerable to existing culture and perceptions about current activities and organizations. In our case, this was a major difficulty to confront. We expected that it was a safer way to real inventions by using an empirical approach based on testing contrasting hypotheses, given that there were well-defined criteria and a goal-seeking strategy for action and development research.

The succeeding empirical exploration and the resulting formal model of healthcare planning for long-term care.

The empirical exploration began with a period of passive naturalistic observations of self-governed development work by the personnel at four units of long-term care. This first period was followed by some action-directed support by the research team according to a management-by-objectives strategy. In addition, a group of three new units was observed where the healthcare system was introduced and organized in the conventional top-down manner.

All our observations were interpreted relative to the chosen theoretical framework and contributed to the resulting model of healthcare planning illustrated in Figure 2. The model was further specified as a flowchart of information processing about the patient and the effects of treatment. Formally, it is a finite-state model with stages of information processing and decision making (Ashby, op cit; Miller, Bush, & Galanter, 1960;  Miller & Chomsky, 1963). Incidentally, a later project on inspection of work environment management used a similar model (RD9).

Figure 2. Model of individual healthcare planning (Strangert, 1983).

The individual healthcare plan is structured as a flowchart of one long-term cycle of diagnostics and action planning, and another, embedded daily short-term cycle of treatments and observation of effects.

The diagnostic stage starts when a decision (b_v) is made to collect (or revise) information (O_b, O_at) about a patient’s problem (P, a disturbance). The information is categorized, interpreted (s), and cumulated progressively as a consequence of incidents and recurrent evaluations. Each diagnostic cycle is associated with a stage of action planning, including goal formulation (g_m) and decision (m).

The main diagnosis and action plan (b_a) are in turn the basis for the daily cycle of treatment (g_a, a) and observation of effects (o_r). The ongoing work is regularly or incidentally followed up, and the cumulated results (r) are evaluated (u) as part of the long-term cycle of individual healthcare plans.

The operational specification of the model necessitated much active empirical work and methodological problem-solving. Some essential methodological challenges will be reviewed in a forthcoming paper.

What can be done better today?

The theoretical perspective based on cybernetics and cognition is not out of date. On the contrary, the idea of uncertainty reduction as means of knowledge acquisition is more promising than ever. It acknowledges that uncertainty is a major characteristic of complex matters, especially human beings and artifacts. In the present case, respect for patients involves both an understanding of the complexity of the individual patient and a serious ambition to develop one's capability to reduce the uncertainty about patients’ needs and disturbances.

However, the earlier theoretical and methodological means of treating intraindividual and interindividual diversity were limited. Specification of complex patterns of individual qualitative characteristics required much empirical scrutiny. Likewise, qualitative inter-individual comparisons became laborious. As a consequence, uncertainty could only be specified in either abstract and general theoretical terms or by meticulous qualitative descriptions of individual cases. The gap of specification between the model prescriptions and the empirical description of individual cases was large.

Nowadays it is possible to use qualitative modeling to simulate cases and outcomes of treatment. This can be a strong important link between generic conceptual modeling and qualitative case descriptions. Thus, the gap between models and authentic individual specimens can be bridged by theory-generated simulations of cases.

References

Ashby, R. (1956). Introduction to cybernetics. London: Chapman & Hall.

Aulin-Ahmavaara, A. (1979). The law of requisite hierarchy. Kybernetes, 8: 259-66

Miller, G.A. & Chomsky, N. (1963). Finitary models of language users. I Luce, Bush, & Galanter (Eds.) Handbook of mathematical psychology, Vol II. New York: Wiley.

Miller, G.A., Galanter, E., & Pribram, K. (1960). Plans and the structure of behavior. New York: Holt.

Strangert, B. (1983). Organisationsutveckling i långvården. Umeå universitet. Forskargruppen för kommunikationspsykologi.

References on this website:

M6. Argument för reglerteoretisk ansats vid utveckling av vård och omsorg (2015)

RD9. On forming perspectives for innovative project planning. Examples from projects on supervision of work environment (2015)

RD13. Perspectives on projects for R&D: some issues to be contemplated (2015)

RD14. On general sources of uncertainty in project planning (2015)