Bo Strangert (RD21)

Action Research. Remarks on its theoretical basis

What is Action Research?

The answer depends on whom you ask. There are quite different meanings, albeit one common defining quality is that 'Action Research' (AR) applies to systematic interventions in real practices, especially social activities and systems. One essential source of confusion is the divergent meanings of the term 'research' in different kinds of approaches. Another common meaning of AR emphasizes the collaboration between research agents and practicians in changing a system (cf. Whyte, 1991; O’Brian, 1998; Chevalier & Buckles, 2011).

There is a muddle of action-oriented approaches that assert some kind of market-led image of the need for knowledge and assistance in social practices and organizations, though many projects seem to fall short of any credible scientific agenda. This may sometimes be true also of approaches that have well-intended purposes to empower weak parts at workplaces by participation in action projects.

Then there are questions about scholarly advances of AR within social science. Commonly, a dominating trend during the last decades is to endorse a critical view of conventional social science because of its ”inability to form questions of practice or social action” (Small, 1995). Besides the research agenda, Small and many others also survey crucial aspects of epistemology and methodology. Their critical questions involve forms of knowledge, the role of ethics and values, and objectivity in social science. Furthermore, the criticism includes data collection strategies and the alleged neglect of the role of the researcher and participants in the research process. O’Brien (1998) has made an extended review of different action research strategies in use.

I certainly agree with the characterization of conventional social science as presently rather incapable of analyzing social practices in-depth and of intervening effectively in complex change projects. Nevertheless, I do not think that some of the suggested deviations from basic principles of science will improve the conditions of practicability. On the contrary, the complexity and diversity of social phenomena and practices are not mastered by relativism and subjectivism in a post-modernistic spirit. The advance should be shaped by theoretical and methodological developments congruent with principles as developed in modern cognitive and organizational sciences. Admittedly, it is a nuisance to use technical scientific language in participative discourses with practicians. On the other side, when discourse about complex phenomena is restricted to common sense and language, it will run the risk of being opportunistic and superficial.

The challenges of participative action research will not be treated further in these notes. (Some formal aspects of project management involving experts and end-users are discussed in RD7 and RD12 on this website). The following lines attempt to give a synoptic outline of some of my scattered reflections about formal AR methodology on this website. I intend to illustrate some major methodological challenges for effective social action research and make a few cursory remarks on possible strategies for advancement. The arguments lean towards recent directions in the philosophy of science on complexity theory and causal mechanisms. The case examples come from early projects consonant with the new directions.

On complexity and diversity of social practice and action

It is a common view that systematic changes in social practice are a complicated business. A loose characterization of a phenomenon like social practice would include a set of many different components (e.g., people, environment, tasks, activities, rules, norms and values, thoughts and reactions) and a set of complex relations between those components.

Figure 1 is a simple outline for the following remarks on how to initially make a conceptualization of a complex social system. It is postulated that the target system exists in reality and needs thorough revision and change. The revision is triggered by the feedback of disturbances or by demands on revised goals. That suggests a reconceptualization based on analysis of system behavior and explication of goals. The term 'Judgment' means the task of processing information that links goals and conceptualizations to feedback and possible design actions.

Figure 1. Outline for the present remarks about the conceptualization of social systems.

A reasonable way of conducting a project requires some capable decision-makers, e.g., proficient change agents participating with representatives from the target system. I will apply a scientific mode to explore the demands on how much change agents could come to grips with complexity issues.

The particular scheme of conceptualization is invented by Bechtel and Richardson (2010). They used Simon’s (1969) notion of "bounded rationality" as a key to understanding human reasoning about complex phenomena and suggested that two heuristics, decomposition, and localization, can characterize how scientists attempt to analyze phenomena in biological and cognitive sciences. They noted that the aim of pure research in these domains during the last decades has shifted from the discovery of general laws or principles to the localization of specific mechanisms, involving complex relations.

The argument of Bechtel and Richardson that the use of the two heuristics depends on human beings’ limited capability of information processing certainly applies also to how change agents analyze, plan and design social practices and systems. That is the premise of the following discussion.

In addition, the necessary interaction between research agents and practicians makes the project still more troublesome in certain respects. Incidentally, Simon’s depreciation of social interaction and judgment in design issues has been addressed critically in later research (e.g., Hatchuel, 2002; Huppatz, 2015).

Briefly summarized, Bechtel and Richardson argue that the analysis of a phenomenon begins with the segmentation of a system from its environment in a way that conforms to its main functions, tasks, or goal. This first step is suggested by collateral theories and facts and is the basis for answering the question of whether the segmented system could be the locus of control for the target phenomena.

An affirmative answer leads to the inquiry of whether analytical decomposition and localization are appropriate. Given positive efforts to decompose the system into components, the next analytical question is:

– Can different system activities be localized in different components?

An affirmative suggests first-order independence between components, while a negative answer leads to a new question:

– Should the description of the phenomenon shift to a lower level?

A negative answer suggests that the analysis should be extended to include first-order interactions between components at a high level, while an affirmative answer suggests reanalysis, leading to ”reconceptualization of the phenomena to be explained in terms of the underlying lower-level mechanisms" (Bechtel & Richardson, 2010, p.195).

The analytical scheme by Bechtel and Richardson seems constructive for analyzing how social practices and systems are designed and developed, especially because the social aspect complicates the definition and decomposition of a phenomenon in two interrelated ways. First, there is an intrinsic difficulty due to the complexity of a social phenomenon per se; a question is whether the applications in biological and cognitive sciences are relevant. Secondly, the design and development process must usually be conducted through participative decision-making in a complex context including different points of view.

Notes on strategies for modeling complex social practices and systems

Which are the practical implications of using the heuristics of Bechtel and Richardson in consultation and applied organizational and social research? Are there some guidelines to avoid being trapped in fuzzy and linear reasoning?

I will use examples from two different domains to discuss the following steps of modeling (concept development) in projects on complex social systems:

• how to segment a system from its environment so the requirements of an open systems approach and a locus of control are fulfilled,

• how to select appropriate components and relational structures for relevant decomposition of practices and systems,

• how to localize ”mechanisms” (i.e. basic operations and processes

On the use of facts or theory to define and analyze target systems for AR

Examples of modeling healthcare systems

These examples of projects aimed at ”improving patients’ need of individual and holistic long-term healthcare by advancing the personnel’s capability of planning and conditions of work”.

We use the outline of Figure 1 to compare two extreme ways of initiating a conceptualization procedure.

1. Exploit facts about a whole existing system. The first alternative is a bottom-up approach and starts by collecting facts about an existing type of healthcare unit for long-term care and its activities with patients. The primary conceptualization then builds upon inferences about how the system, its locus of control, functions, and tasks are defined and conceived at present. Commonly, the description uses conventional terms of organizational attributes and tasks. The next step will be to analyze the consequences of any adverse feedback or demands on goal revision. The outcome may suggest some major or minor revision of the existing conceptualization. It is done by collateral theories which have implications for possible actions and redesign. 

Selection and analysis of an existing prototype system can ensure realism and inclusiveness of necessary components and relations. A very strong argument is that different components are integrated according to facts and experiences. That sort of ”natural” point of departure for possible remodeling would certainly be accepted as trustworthy and understandable by most involved parties.

Notwithstanding the obvious inclusiveness of relevant features and a representative and established locus of control, those very characteristics are probably the strongest hindrances to an unbiased development process. Inbuilt qualities of management and routines are major threats to renewal. For example, well-established qualities of the participating administrative, economic and medical professions would initially and easily impose their particular restrictions on the project process. That would be a nuisance if, as in the present case, we realized the need for fresh thinking about the complex challenge inherent in the project goals.

A moderate version of the bottom-up approach to an existing system would restrict conceptualization to selected parts or functions of the system because of clues from feedback information or demands on goal refinement. This is the most common strategy of modeling and presupposes a decomposition into existing subsystems and components as a basis for selecting certain target functions for an independent analysis. It is assumed that separate analytical solutions of development can be integrated later into a coherent model for the whole system. In general, the risk with such an additive strategy is that assumptions of independence and a linear combination of subsystems would be inadequate in complex cases.

2. Construct a preliminary model to steer empirical investigation and inventions.

A contrasting strategy is to formulate a new tentative formal model of locus of control, based on the prime goals of core processes (e.g., ”Advancing individual and holistic long-term healthcare”). This was the strategy used in a case study (RD15-18 and M6-8). It meant the choice of a control-theoretic model for framing two interacting subsystems: one of the individual patient and another of a professional healthcare unit. The model constitutes theoretical assumptions of necessary conditions for the core processes, that is, an initial theoretical guideline for decomposition and localization.

The decomposition focussed on internal as well as interrelated essential functions of the subsystems. The patient system is the root of a superordinate hierarchical control structure, and it presupposes interactions between internal physical, psychological, social, and basic behavioral (ADL) functions and components for satisfying vital needs and coping with disturbances.

The professional subsystem will act to support the patient’s need satisfaction and coping capability. Support functions comprise control cycles of diagnosis, action, and follow-up against individual criteria on needs and disturbances. Consequently, the continued decomposition involves the allocation of control functions to operations, including real tasks, procedures, and educated personnel.

The elaborated control system includes operations for planning, decision-making, action, and follow-up. Figure 2 shows an example from the project (RD17). Note that this preliminary model presupposes a network of interactions between healthcare agents and the patient, including many data sources and specific transactions localized on a lower and deeper level of analysis.

Figure 2. Flowchart of a control model structure for planning, decision making, and conductance of professional healthcare operations (see RD17 for details).

The flowchart shows only closed healthcare operations of the professional subsystem. Its direct and localized low-level relations to the patient are not specified, and neither is its high-level association with its administrative context, society, or research. As mentioned, it is an abstract model for guiding further concrete structuring and low-level localization of core processes for healthcare. Many corresponding basic mechanisms of medical, physical, and psychological categories can where applicable, be localized given an adequate model and patient participation.

The case study had a rather elaborate stage of top-down conceptualization before facts from healthcare practices were included. This can be regarded as a serious risk of missing crucial characteristics of the real phenomenon. However, that is only a risk when irreversible and unconfirmed decisions and actions are made based on preliminary conceptualization. From some critical point in time, it will be necessary to begin empirical testing of basic conceptual propositions. This signifies the onset of the action or intervention ingredient, which in any case and time should be hypothesis-controlled.

Our strategy of AR in the case was to combine top-down concept development (CD) with experimentation (E), that is, interventions in practice. The purpose is of course to test the modeling continuously. However, testing cannot be done without considering the factual characteristics of the phenomenon, which in most cases includes already existing healthcare activities. Therefore, action research requires a concomitant evaluation of existing systems, that is, by attempts to conceptualize their structure through decomposition and localization. The outcome of the conceptualization is then contrasted with the preliminary formal model, constructed based on independent and comprehensive goal analysis.

Apart from the initial phase of constructing a preliminary top-down model, the further conceptualization and evaluation of existing systems progress concomitantly. The resulting conceptual contrasts between the constructed model and the evaluated existing systems form the ground for continuous planning of hypothesis-controlled empirical interventions.

As Bechtel and Richardson argue, progress in science is characterized, first, by the detection of faulty decompositions and localizations of phenomena and, secondly, by replacing them with new and better conceptualizations through a cyclic process of hypothesis formation and validation.

However, a total and definitive explanation (conceptualization) of really complex social activities is not possible even in a very long time perspective. It could not be congruent with the meaning of 'complex' as multifaceted, dynamic, probabilistic, and partly non-transparent. Therefore, the strategy should settle for a process of continuous exploration and improvement.

This progressive view concerns the complementary issue of how to develop and change social systems. It means a focus on purposeful action and design. Before looking at the action perspective, another example of the choice between conceptual frames will be summarized.

The preceding example contrasted a truly theoretical model with a conceptualization of a typical system in practice that was inferred based on facts. Another kind of contrast is between alternative theoretical models for guiding development and design. This type of contrast is about choices of action alternatives and design options.

On the choice of contrasting conceptual frames for Action Research

Contrasting different project models can be seen as a parallel to ”the plausible rival hypotheses strategy”, which Campbell (2003) described as a plan of contrasting alternative hypotheses with extended networks of implications that can never be completely examined. Consequently, the conclusions and decision-making about alternatives must often be based on many different and specific indications over time regarding particular contexts. The view pertains to 'conceptualizations' (models) of complex phenomena which include extensive networks of implications.

Examples of modeling work environment supervision

The three selected examples of modeling a complex system for supervision show consequences for action research of choosing a particular conceptual frame. A detailed background to the case is presented in RD9 and T19. The description is here limited to a few distinguishing theoretical points.

The case originated from the Swedish Work Environment Agency’s projects of designing procedures for supervision of ”Systematic work environment management” at workplaces during 1990-2000.

The projects were founded on the brand new provisions on ”Internal control of the working environment”, IK (AFS 1996:6), and the subsequent provisions, ”Systematic work environment management”, SAM (AFS 2001:1), which contained binding rules for ”the work done by the employer to investigate, carry out and follow up activities in such a way that ill-health and accidents at work are prevented and a satisfactory working environment achieved.” (SAM definition, AFS 2001:1).

Provisions stand for a general system of connected rules for systematic work environment management that can be implemented at all kinds of workplaces. Each responsible employer is supposed to be able to implement the system.

The relation of the particular systematic work environment management (SAM) to the ordinary management of activities is stated in SAM as follows: ”Systematic work environment management shall be included as a natural part of day-to-day activities. It shall comprise all physical, psychological, and social conditions of importance for the work environment.” (AFS 2001:1 Section 3)

General Recommendations are issued to facilitate the practical application of the rules:

”General Recommendations have a different legal status from Provisions. They are not mandatory. Instead, they serve to clarify the meaning of the Provisions (e.g. by explaining suitable ways of meeting the requirements, giving examples of practical solutions and procedures) and to provide recommendations, background information, and references.” (AFS 2001:1)

The essential General Recommendations to guide the basic Provision Section 3 about the integration of SAM to the ordinary management of production, economics, and quality include:

”Work environment issues need to be handled within the activity in the same way as production, economics, and quality, not as a separate system. Many operational decisions have a bearing on working conditions, and the consequences for the work environment, therefore, need to be assessed and considered before the decisions are taken. …”

”The employer can also apply voluntary systems of quality assurance and environmental management. Systematic work environment management can, where appropriate, be coordinated with these systems, but it is important that, when thus coordinated, work environment issues should continue to receive sufficient scope and attention, e.g. as regards questions concerning musculoskeletal ergonomics, psychosocial conditions, job modification, and rehabilitation. …

”The employer needs to take into account all factors potentially impacting on the individual persons’ work situation. This does not only mean things capable of negatively affecting health and safety. A good work environment contributes towards good health and means more than the absence of illness and accidents.” (Excerpts from General Recommendations: Guidance on Section 3)

Supervision. The task of designing an inspection procedure for supervising the implementation of SAM (IK) should consider the intended purpose of the Provisions. For example, the set of provisions can be interpreted as conditions of a control structure for work environment management, including a work environment policy, risk assessment, measures to be taken, and follow-ups. This necessitates a set of appropriate inspection procedures – that is, a control system of higher-order, ”control of control”. The design of such a system can then begin with preliminary conceptualizations that either conform closely to previous norms of inspection procedures or opens up for reassessment.

The previous papers on the subject (RD7-9, T19) give a comprehensive picture of three contrasting frames of conceptualizing an approach to supervising practices of IK or SAM. A summary, couched in the terminology of Bechtel and Richardson, is as follows:

Any conceptual frame should direct the design of a system to include a locus of control for supervision. Since the object to be supervised is also regarded as a control system, a core question will be how the interaction between the two systems – the inspection authority and the workplace management – should be modeled.

The reasoning so far in terms of control theory has important consequences for the selection and design of components and localization of activities within and between systems (see RD9 and T19 for details).

The regulatory framework of Provisions (and the superordinate Work Environment Act) is of course a set of common mandatory components for both systems. Provisions, singly or in combination, should be key components in all judgments and actions of employers and inspectors.

Yet, their role in designing inspection procedures may differ depending on how the locus of control is modeled.

In one of the examples of conceptualization, called the Administrative or Regulatory model, the prime purpose of control is rule compliance. The inspection of Provisions is localized into specific activities of survey, judgment, feedback, and could-be sanctions. A corresponding localization of rule-governed activities by the employer would occur at the workplace. Thus, the outcome of inspections of SAM (IK) concerns the employer’s rule compliance as evaluated by the inspectors’ specific rule-centered procedures.

The localization of control is on forms and rules as the basis for systematic work environment management. The application is general and not mode-specific; types of risks and effects are optional and secondary to the administrative forms of managing them. The corresponding inspection procedures shall match this view. The success of implementation is assumed to depend crucially on how Provisions are formulated and communicated to employers and employees. Hence, much effort should be spent on devising acceptable and comprehensive inputs of information to guide the target groups.

The purpose of another type of conceptualization, called the Effect-based model, is primarily to inspect the risks and negative effects of the work environment to judge and force the employer to cope with possible and existing problems according to relevant Provisions. Thus, its locus of control implies that risks and negative effects shall be detected, causes can be inferred and counteractions conducted. The inspection relies commonly on and primarily on data from surveying employers and employees about indications of risks and ill health and damages at work, even though the need to reinforce positive work aspects should be a task, too. Focus is frequently on psychological and social issues, and the communicated information about effects is usually subjective. The conclusions from surveys are shared among parties and directed to the employer to manage according to SAM (IK).

The third type of conceptualization, called the Process-based model, contrasts markedly with the preceding two regarding its locus of control. Here the goal is to induce a management process to integrate the management of production, economics, and quality with work environment management, the qualities of which are assumed to be mutually dependent. Any perception, feeling, decision-making, or action of the employer and employees may influence all activities and events at the workplace, and conversely, also be influenced by any event at the workplace.

The key implication of this model follows from Section 3 of the Provisions (2001:1): ”Systematic work environment management shall be included as a natural part of day-to-day activities. It shall comprise all physical, psychological, and social conditions of importance for the work environment.” Consequently, the main focus of supervision should be on the process of integrating work environment considerations into daily production management.

Figure 3 illustrates some distinctive differences between the three models. The administrative (regulatory) and the effect-based models of supervision highlight the separation between the systems of 'Work Environment Management' and 'Production Management. ('Production management is a synonym for management of production, economics, and quality.) The purpose of ”natural integration” is regarded as the employer's duty to establish. It can only be indirectly reinforced by inspection through the requirements of rule compliance (as in the administrative model), or through the feedback of WE-effects which are linked to appropriate Provisions (the effect-based model). Hence, these two models share an underlying assumption of linear (additive) relation between the management of the work environment and of production. A possible cognitive ”mechanism” of natural integration would be superficially localized to the will and mind of the employer when confronted with provisions and inspection activities. But such an explicit mechanism is neither formally stated in the models nor inferable from the inspection procedures.

Figure 3. Three schematic control models of supervision.

Conversely, the process-based model assumes that the requirement on Work Environment Management as a natural part of day-to-day activities is crucial and should be inspected directly. This is illustrated in Figure 3 as the inclusion of work environment tasks in the total management activities. The assumption of interaction is theoretically localized as cognitive mechanisms of the employer which should be reinforced actively by inspection. It is also assumed that the reasoning of employers and employees is predominantly based on matters of work production – a primary mental framing that needs to be transformed to include work environment aspects. The strategy of inspection should guide the employee and staff towards ways of conforming to this general requirement in all situations where both production and work environment qualities ought to be congruent.

A related contrast is that the process-based model differs from the administrative and effect-based models regarding the possibility of separating input/output factors for work environment and production. The process-based model assumes that many production and work environment factors interact and therefore are difficult to separate analytically and practically.

On consequences for continued CD&E

and implementation of supervision

Different early conceptualizations, as in the present examples, would lead to divergent consequences in the continued analytical and empirical endeavor. Although there will be a subset of common components of norms and routines for inspection, the levels of decomposition and localization of mechanisms could differ considerably between project models. The consequences would manifest explicitly during the design and implementation of inspection procedures.

For example, the regulatory model permits the simple transformation of regulations into well-known components of tasks, instruments, and procedures according to the praxis of supervisory authorities. The localization to conventional control circuits can be direct and objectively defined in terms of formal routines and restricted judgments of subject matter. Though it may be a disadvantage for some matters that qualified judgments can not be made. On the whole, the implementation of inspection practice would look straightforward from the point of view of authorities and senior inspectors.

The strength of the effect-based model is its emphasis on observation and judgment of effects, notably of a psychological and social kind. However, this makes decomposition difficult because instruments and data involve subjective judgment and special inspection capabilities of handling social interaction at workplaces. So localization of psychological and social ”mechanisms” must be made subjectively and based on complex and situated information. That complicates the process of feedback and getting management to act constructively.

The process-based model locates the control mechanisms on a deep level of perception, decision-making, and actions of the employer and staff. It implies that work environment management is about a complex and multifaceted task that ideally should be closely adapted to local conditions and goals with recurrent supervision over time. It also requires highly competent inspectors who are familiar with the particular type of activities at each workplace; this is implied by the postulate that production and work environment issues should be dealt with as integrated tasks. These strong but realistic preconditions for work environment management raise demands which can be difficult to realize in practice due to lack of resources and mastering.

Hypothesis-controlled action

A relevant AR should involve hypothesis-controlled research based on conceptualization (modeling) and experimentation (e.g., systematic interventions in social systems). The focus will now shift towards questions on causality and how to make things happen (Woodward, 2003).

As noted in the introduction, there are purposeful action projects at workplaces that engage people to participate actively in improving work conditions. It can be a process of trial and error led by participatory decision-making with little or nothing about formal conceptualization. The possibilities and problems of participation in organization development are not at issue now. The focus is on scientific foundations for action research methodology.

Hypothesis-controlled interventions in complex contexts give rise to more methodological requirements than can be met by true experiments or exclusive quasi-experimental methods. (Experimental procedures with randomized control are of course seldom appropriate in practice.) The difficulties are associated with the multifaceted character of social practices. Any intervention can generate a huge array of consequences for the target object, and many interventions are inherently complex. Well-known facts and experience can suggest some possibilities or prevent mistakes, but complex problem solving needs conceptual and methodological instruments that advance innovation. 

The present view of hypothesis-controlled intervention can be outlined with the aid of Campbell’s (1983) ”plausible rival hypotheses strategy”, as mentioned before. Applied to the examples of three contrasting models of supervision, it is easy to conceive of hypothetical interventions generating different but partially overlapping networks of implications; the critical differences should be related to the distinct control structures of supervision illustrated in Figure 3.

All possible implications of the models cannot be tested, nor is it necessary. It suffices that distinct patterns of implications are derived from contrasting hypotheses stemming from the models. The pattern of implications shall be possible to translate into empirical indications or operations, which are reliable and verifiable or valid representations of their conceptual counterparts.

Some examples from the case of supervision may clarify the set of steps to be taken:

The administrative/regulatory model can generate hypotheses about implications of rule compliance for single or sets of Provisions (SAM or special ones). The requirements of rule compliance shall apply to various tasks, activities, and environment qualities at workplaces as well as to specific procedures of inspection.

Rule compliance is formally a simple general principle for control and hence a very useful key to the administrative application. Yet, it requires some extra efforts to translate the principle into specific criteria on what rule compliance means concretely at each workplace and how it shall be inspected. If that is attainable, the use of SOPs can provide benefits of time and costs for both management and inspection. However, indiscriminate use of SOPs can cause simplistic and static judgments and measures.

The effect-based model generates preferably hypotheses about implications of work environment risks and corresponding management actions. It is frequently and particularly applicable to situated psychological and social workplace problems, which require social interactions between inspectors, employers, and employees. In addition, the transactions during the phases of survey, feedback, and action put special demands on domain-specific competencies of the inspector, e.g., knowledge of possible cause-effect relations and the ability to properly handle backward reasoning about fuzzy problems.

Thus, the demands of this approach contrast radically with the administrative model and its weak and distant relations to real effects. From a theoretical view, on the other hand, the effect-based model, as well as the administrative model, treats the systematic work environment (SAM) as a general system that can be flexibly added to the inherent production system. Although couched in terms of becoming a ”natural part” of the daily activities, SAM does not include any formal statement about how ”the mechanism of integration” should be represented other than as a simple concatenation. Consequently, it is fair to localize the integration mechanism theoretically as a linear high-level relation between two large components. In practice, the real integration is left to a process of trial and error, carried out by the responsible employer and staff and superficially guided by Provisions and General recommendations.

The process-based model differs from the other two models because its locus of control concerns the interaction between production management and systematic work environment management (SAM). However, such a statement of first-order interaction at a high level lends little guidance when hypotheses on supervision strategies will be formulated. Accordingly, the conceptualization of natural integration must be shifted or reduced to lower levels to derive hypotheses about supporting inspection activities.

Cognitive modeling of problem-solving is a self-evident way of conceptualizing the management process of integrating production and work environment matters. For the employer, it is necessary to infer the consequences of decisions about production on the work environment and vice versa. Incongruences must be identified and eliminated. For example, an appropriate control strategy is to select well-known problems at work and analyze their possible causes and consequences from both production and work environment perspectives. By comparing the consequences with relevant multiple goals and standards, appropriate alternative actions can be searched and decided upon.

In brief, the control-theoretic stance of the SAM provisions can be used both as a model guide for application by employers and for inspection. Its advantages are based on its capability of specifying hypothetical implications of decisions and actions more accurately than the other two model approaches. Its practical disadvantage lies in the relatively high demands on inspectors regarding both analytical skills and expertise in production matters.

Notes on causal reasoning and intervention

There is a lot of factors that may be characterized as 'causal' to work environment effects. Some are conceptually well-defined with clear extensions to physical matters. Other factors, especially about the organizational and social work environment, are very complex and have imprecise semantical definitions and fuzzy referents in reality. The Swedish Environment Authority has recently issued Provisions (AFS 2015:4) about how to systematically manage the latter types of work environmental factors; they are subsumed under the categories of Workload, Working hours, and Victimization. This move strengthens the possibility that psychological, social, and organizational problems can be dealt with more effectively. The demands on competent handling will be more explicit, but the causal relations and desired actions within these areas are still complex and difficult to manage and supervise.

In addition to factors at the workplace, there are miscellaneous indirect influences on work environmental effects from society and private life. It is not realistic to expect that the causality of such possible effects can be determined unequivocally. That is, a description of a single or a few conditions may not be sufficient to explain a subsequent state of conditions – the explanandum does not follow from the explananda. Advanced analysis of a phenomenon probably reveals that multi-causality, interactions, and feedback circuits are the rule rather than the exception.

Accordingly, the business of validly explaining working environment effects through rigorous causal explanation is problematic from a scientific point of view. This conclusion pertains clearly to the organizational and social work environment. However, Elster (2007) argues that causal explanation has a given pragmatic role in social science. His view resembles Campbell’s idea of formulating and testing rival hypotheses regarding causal implications. The conceptual implications of the favored hypothesis should be corroborated by observation of facts (data; i.e., be physically realizable).

Woodward (2003, p.10) suggests that ”…our interest in causal relationship and explanation initially grows out of highly practical interest human beings have in manipulation and control.” Thus, information about invariant relationships can be used to manipulate conditions. Conversely, the notion of manipulation can even be used counterfactually to explain a phenomenon that cannot be physically manipulated, for example, by reasoning that if manipulation of some hypothetical causes were possible, then the phenomenon would be affected.

Woodward’s ”manipulability conception” as a key to causal explanation is in line with recent views on how technological applications have impressively influenced the development of science through experimentation and instruments. ”Making things happen” is in principle a constructive way of investigating and changing complex phenomena. The strategy is closely related to 'design' as the construction of a set of innovative means-end actions to attain a useful and research-based service or product.

Applied to social practices, such an approach may somewhat resemble the old and wider art of ”social engineering”, which fell into disrepute partly because of the lack of constructive theory and methodology at the time. Incidentally, Popper (1945) examined its possible democratic use in society and proposed that a ”piecemeal social engineering” based on scientifically validated steps should be acceptable. More recent and well-known examples of scientific design applications to psychological and social practices include the work of Simon (1969, 1996) and Carroll (1997).

A summary of Woodward’s (2003) intervention strategy is that a hypothesis 'C causes E ' is tested by examining what will happen to E through an intervention I on C. The intervention I on C is itself causal, and this relationship needs to be justified independently from the claim of C causing E.

In cases of social intervention, there is of course a large set of specific auxiliary assumptions about the intervention and its influence on C. In addition, there may be many other still unnoticed conditions associated with I. Hence, the use of I for causing C may probably be uncertain and fallible due to deficient or inappropriate experiences and incomplete inferences. The influence of likely preconditions must therefore be carefully evaluated during each step of intervention. It means scrutinizing a pattern of qualitative I-C-E indications for proof.

However, although there is a complex causal structure behind an intervention, the analysis need not be developed beyond a point of practical utility. What this means for a given practice is an open question. The kind of reasoning can be briefly illustrated regarding the example of supervision.

Supervision as an intervention in

systematic work environment management

On a rather high and rough level of analysis, the three models of supervision generate hypotheses about the effects E of separate types of supervision Ix (rule-based I, effect-based I, or process-based I) on C (systematic work environment management), i.e., Ix→C→E, where E represents the intended work environment effects (e.g., on physical, behavioral and subjective states).

A possible indirect causal relation of Ix on E cannot be disclosed empirically without many auxiliary assumptions which are awkward to justify. Furthermore, the supposed intermediate causal relation C→E is difficult to demonstrate due to probable influences on E of a lot of interfering internal and external factors. Consequently, the importance of I on C should ideally be assessed by qualitative observations of specifically connected inspection–management actions. If significant and lasting I→C relations were observed, they could, for example, be interpreted as inspection-induced transformations of the employer’s decision-making and action, or as an effected organizational change of C.

The importance of conceptualizations that generate contrastive hypotheses is necessary when experimental control designs are ruled out. Each hypothesis must include a unique pattern of implications that differs in some crucial ways from alternative hypotheses. The contrast between alternative hypotheses of interventions should have decisive consequences for the supposed cause-effect mechanism C→E. Operationally, the impact of interventions I could first and directly be observed through indications of C (SAM) and, secondly and indirectly, be inferred through indications of effects E (work environment effects).

For example, the locus of control for the rule-based model is about ”mechanisms” for rule compliance. Its level of description is general, and it can easily be applied to most components of the SAM system. However, its impact depends on how well the indicators of the inspected SAM components match the corresponding intended properties of the rules. If a connection between an indicator and a rule is superficial or casual, its supposed causal impact though compliance will be weak. In addition, the explanation of such an indication is uncertain.

In conclusion, the effectiveness of rule compliance depends on the clarity of rules and their implementation. This model does not specify the control mechanism sufficiently well to make it possible to infer or observe if SAM is integrated into customary management. Therefore, an implementation can be described and tested as an additive combination of SAM and production management at most.

A similar linear assumption of implementation would apply to the effect-based model, too. There are no specified assumptions concerning how SAM complements customary management of production. However, the model is determinate in comparison with the rule-based one, because in this case SAM is based on the assessment of content-specific risks and effects. That would make it easier to plan and effectuate relevant actions unequivocally, especially for indefinite psychological and social disturbances. The power of more determinate assessment and counteraction is nevertheless dependent on the unresolved capability and determination of the employer. It is difficult for an inspector to judge those preconditions because this model includes no specifications about how to integrate SAM and production management.

The process-based model contrasts with the foregoing ones by stipulating the requirement that SAM shall be integrated into the production management, following the Provision Section 3 in SAM (AFS 2001:1). It means that hypothesis-controlled interventions, i.e., inspections, should enforce the implementation of procedures and measures to ensure integrated management. Inspection is for this reason a necessary transactional complement to the Provisions and General Recommendations. The hypotheses behind process-based supervision include a set of specific implications for successful implementation and judgment of integrated management. The specifications should theoretically and practically involve conditions about cognitive and communicative aspects of decision making and action.

The need for elaborate strategies for action research and development

As the examples illustrate, complex social practices put high demands on theoretical and methodical approaches to explanation, development, and change. This standpoint may invoke resistance from employers who are afraid of losing control of a change process and therefore want to preserve the status quo or at least conduct changes in simpler and transparent steps. A different type of reaction to scientific-based action research comes from planners who prefer to design a complete blueprint for the end product and its construction before the real design and implementation process.

How to handle resistance towards organizational and social change is not a subject of this paper. However, it would be risky to succumb to the lure of a generous project invitation which contraindicates important action research conditions. The message is that complex issues should be managed by the use of adequate theoretical and methodical strategies.

An ideal strategy for the development or change of complex practices involves a long-term endeavor to generate continuous and sustainable improvement. It implies a dynamic open system-theoretic strategy instead of either bottom-up approaches or readymade designs that tend to lock up the change process by relying on faulty experiences or overrated capability. The strategy should use contrasting conceptualizations that generate hypothesis-controlled action as well as testing of divergent implications to infer intended purposes. Using rival models and hypotheses is a compelling substitute in discrete case studies when experimental control designs and comparisons are impossible. The selection of action or design should be based on that model or hypothesis which generates the most constructive implications in fulfillment of the project goals.

References

AFS 1996:6. Internal control of the working environment. Swedish National

  Board of Occupational Safety and Health.

AFS 2001:1. Systematic Work Environment Management. Swedish Work

Environment Authority.

AFS 2015:4. Organisational and social work environment. Swedish Work

Environment Authority

Bechtel, W. & Richardson, R.C. (2010). Decomposition and Localization as

Strategies in Scientific Research. Cambridge, Massachusetts:

MIT Press

Campbell, D.T. (2003). Foreword. In Yin, R.K. Case Study Research. Design

and Methods. 3rd ed. London: Sage

Carroll, J.M. (1997). Human-Computer Interaction: Psychology as Science

of Design. Annu.Rev. Psychol. 48:61-83.

Chevalier, J.M. & Buckles, D.J. (2011). A Handbook for Participatory Action

Research, Planning and Evaluation. Ottawa, Canada: SAS2

Dialogue

Elster, J. (2007).  Explaining social behavior. Cambridge University Press.

Hatchuel, A. (2002). Towards Design Theory and Expandable rationality:

The Unfinished program of Herbert Simon. Journal of

Management and Governance, 5. 

Huppatz, DJ. (2015). Revisiting Herbert Simon’s ”Science of Design”.

Cambridge, Mass.: MIT, Design Issues: Volume 31, No.2.

O'Brien, R. (1998). An Overview of the Methodological Approach of Action

Research. www.web.ca/～robrien/papera/arfinal.html

Popper, K.R. (1945). The Open Society and its Enemies, Volume I. London:

Routledge & Sons

Simon, H. (1969/1996). The Sciences of the Artificial. 1st/3rd ed.

Cambridge, Mass.: MIT Press

Whyte, W.F. (1991). Participatory Action Research. London: Sage.

Woodward, J.  (2003). Making Things Happen. A theory of Causal

                                  Explanation. Oxford University Press

Cited papers on this website:

Strangert, B. (2014). On the choice of ”perspectives” when investigating      

                       complex phenomena (RD7)

  –   (2014). Unclear reasons behind diverse perspectives in the initial project plans (RD8)

 

  –   (2015). On forming perspectives for innovative project planning.

                            Examples from projects on supervision (RD9)

  –   (2015). Perspectives on coordination of a planning process (RD12)

  –   (2015). Model-theoretical reasoning to give structure to complex

                            realities: A case of healthcare planning (RD15)

  –   (2016). Case-based reasoning and qualitative modeling for

                             representation of generality and individuality (RD16)

  –   (2016). On design-scientific approach to healthcare planning (RD17)

  –   (2016). On design-scientific endeavors:  Choice of task perspectives

                             and psychological framework (RD18)

   –   (2015). Argument för modellteoretisk ansats vid utveckling av vård

                             och omsorg (M6)

   –   (2015). Om metodiska konsekvenser av val av perspektiv: En

                     designvetenskaplig tillämpning inom vård och omsorg (M7)

   –   (2015). Kan begreppsutveckling förbättra konsekvensanalyser vid

                     arbetsutformning? Exempel från projekt om vårdplanering

                             (M8)

   –    (2015). Exempel på uppdragsmetodik: att konstruera teoretiska

                   perspektiv på tillsyn av arbetsmiljön (T19)