Gerard de Zeeuw

Universiteit van Amsterdam
Centre for Innovation and Co-operative Technology
E-mail: zeeuw@phys.uva.nl
 
 
 

Abstract

Interest appears to be increasing in the study of values or experiences that are needed to realise more supportive communities, more social societies, more responsible businesses, more serving services and less social violence.
Well-known research methods that were developed over the last two centuries do not seem to help in getting the desired results, as evidenced by the proliferation of 'new' research paradigms, methodologies and even philosophies.
In this paper a number of reasons are suggested and explored why both the standard and the 'new' methods still appear to fail. It is proposed that what is needed are research designs to test advice, rather than observational statements.
 

Introduction

Certain phenomena do not appear to rate discussion as to whether they should be eliminated. Examples are social ills such as conflict and war, crime, corruption, violence against certain groups, bad politics and bad management. A general rejection does not seem to help, however. In every age and time at least some people prefer to do what is rejected. This is not for want of effort to stop this. People have seen fit to establish police, mount movements, teach and give managerial advice-and appear willing to increase their efforts further.

Sometimes they try to identify what money to spend on the police rather than on, for example, education. Sometimes they try to 'understand', and thereby to 'know' how to act. In all cases the criterion is that no new social ills should appear due to the act of resolving earlier ones, or at least that solutions become less and less expensive. Interestingly, the third approach, that is 'understanding' and 'knowing', often proves least effective on this criterion, although its history of technical successes promises the most.

Approaches to deal with social ills usually take the form of advice. This includes utopias, long and short term plans to change societal structures that are seen as inimical, proposals to improve on the legal system, social etiquette, theories and methods for management. It is assumed that the ills to be removed are due to the structure of social organisation, and hence that they can be resolved by advising some or all people to change that structure, that is by addressing them so the desired changes take place.

The difficulty of identifying 'good' advice seems succinctly summarised in a question and a claim quoted from Fischhoff [1]. The question is: "What entitles us to give general advice?" The notion of 'general' is meant to refer both to who is addressed, and to how broad the activities advised are to be defined. The second states a norm: "Yet, one should no more expose people to untested advice than to untested medicine." This suggests an answer. The 'title' to giving general advice derives from having tested the advice.

The aim of this paper is to explore what this answer means, in particular the notion of 'testing', as well as how one might create 'designs' for testing, that is research designs. The first question that has to be raised is what one might test against, or in other words, what 'good' advice would mean. Advice such as 'drop the bomb' is general, but not necessarily 'good'. The second question is what may be used to facilitate a 'test'. This leads to a proposal for research designs-that is forms of testing prior to giving advice.

Quality of advice

Advice will be understood as a process of transmission: someone advises and someone is advised. If the advice is taken, the advisee will do something that s/he presumably didn't think could be done, to achieve something s/he or others desire. Advice may take any form and have any length. It may be built out of parts that function themselves as advice. It may include advice on how to produce advice. It may be simple and addressed to anybody, as well as complex and addressed to one person only.
Usually there is an interest in attributing some quality to advice, and hence to constrain advice in 'quality sets'. Examples include sets the individual elements of which guarantee any or some well-defined effects (with no side effects), or are inexpensive, or address some people and not others, or are characterised by the fact that their advisor is trusted, etc. Boundaries to 'quality sets' may be fuzzy, suggesting that advice may be relatively costly, relatively general, etc.

Advice also may be included in 'quality sets' that are not seen as desirable. One may think of advice through which the advisee achieves something s/he does not like ('give me your money'), or advice that has negative effects for both the advisor and advisee ('drop that bomb'), or advice that is difficult to realise ('be good'). These examples indicate that the quality of advice will depend, at least partly, on what is 'culturally' desirable.

In the same way that advice is constrained in 'quality sets', accepting a particular advice implies that something or someone else is constrained, in this case the ad-dressee or addressees. In more common terminology, advice 'qualifies' the ad-dressee(s) to act in one way rather than another (and implement an intention or 'value'). This implies that s/he is to recognise or observe certain resources, and support a 'personally' desirable act-which in its turn may result from advice belonging to a 'culturally' desirable 'quality set'.

Testing

The notion of testing has a long and convoluted history. As a concept it proved more difficult to clarify than the related notion of (mathematical) proof. Testing was missing from the Aristotelian view. The aim of 'science' was to identify what changed according to its 'nature' or 'soul', and to distinguish this from what changed for other reasons and was 'artificial'. Testing would have been artificial, imposing on the natural, and hence not informative about the latter.

Proving a statement (mathematically) means that it is shown to belong to a (usually) larger set of statements that is constrained by another set of statements (assumptions). The leap from 'proof' to 'scientific test' has been to continue to distinguish the two sets-but to require in addition that both refer to (reported) observations, and that the relation between the two sets is not falsified by, or in conflict with, any of these observations [2].

The difficulty in applying this notion of testing to advice is how to link the latter to observation. Advice usually is in verbal form, and at best urges to observe. Two conditions were suggested, however, to establish the desired link. Firstly, advice is taken to belong to the first (large) set if and when addressee(s) are able to recognise and reduce deviations from their quali-fications. Secondly, advice is taken to belong to the second (small) set if and when addressees are able to recognise and reduce deviations from its 'quality set'.

Taking these suggestions into account, the following concept of 'testing advice' is proposed. An advice is said to be 'under test' if there is an interest in identifying both conditions. It is said to 'have been tested' if it can be shown not to have failed either condition. If some advice is 'new' in that it has not been 'under test' yet, it may be recognised as belonging to a 'tested' set. This is the case if such advice allows for recognition of deviations of this ('quality') set, and if addressees prove able to maintain the qualifications it transmits.

During a cholera epidemic in (19th century) London it was advised to close one water pump [3]. Firstly it had been observed that those who got ill had drunk from that pump (or rather that those not getting ill satisfied the 'qualification' of not drinking). Secondly that the pumps were not linked, which made closing the one pump effective in preventing drinking (and allowed the councillors not to deviate from their 'quality set'). It took valuable time before both conditions were accepted as a 'test' (and the advice as 'good').

Designs for testing

A generic design to 'test' advice may be derived from what was proposed in the previous section. In the case of cholera it eventually proved possible, for example, to split all 'quality sets' into two parts, one part being common to all. This part consists of a description of how a specific organism gets distributed and multiplies in the human body. The other part varies over all 'quality sets' and consists of activities anyone may implement using the descriptions in the first part-for example to stop the disease, or to use it as a weapon.
The first part refers to what usually is considered 'knowledge'. Acquisition may proceed effectively using standard research designs-which we thus surmise to have been developed to allow for the separation into two parts, as indicated. The second part consists of action designs (or qualifications) anyone may implement using the results. The separation into these two parts constitutes a special case, therefore. Results help to guarantee the effect of actions, whoever wishes to use the results in whatever action.

This kind of guarantee obviously is quite advantageous. It also sometimes is deplored. One may think of the criminal use of technology. Interestingly, there also are situations where it is rather the lack of a guarantee that is deplored-for example, when we intend to prevent social violence, or cure or overcome bodily failures. The difficulty is that separating the first part now is possible only by using the second part (making the first part less general). This implies imposing the intentions or values involved.

There are good reasons to avoid impositions. Firstly, there are ethical objections to restricting the freedom not to accept an imposition or dictate. Secondly, there are technical objections. Imposing implies that effort has to be spent to maintain what is imposed, that is to oppose resistance. This suggests looking for 'better' research designs. One obvious alternative is to 'advise' the qualifications involved, and to allow for their self-maintenance, as well as for an 'escape' whether or not they are maintained [4].

An example may clarify the above. Fischhoff [1] collected a large set of advice to women to prevent rape, which he then tried to test using a standard research design. It did not prove possible to identify 'mechanisms' of rape (or generally independent 'first parts'). He therefore tried to separate a possible first part by attempting to qualify women as decision-makers. This implied 'advising' them on one kind of activity, and on one type of qualification, which he hoped to be independent of further forms of knowledge use.

To test this possibility, he acquired observations for each single piece of advice he had already identified, for example on what effects to expect, with what probabilities, and on what selection to make under what priorities. In other words, he tried to discover a 'knowledgeable decision-maker' as a first part, rather than knowledge. If this had been possible, anybody 'using' such a decision-maker would have had to be effective as an actor, or alternatively, would have been able to resist impositions, including those of rapists.

Unfortunately, Fischhoff failed in this attempt, as he admits. The desired effects could not be guaranteed. They continued to depend on other activities and forms of use, in particular those of the rapists. We may also say, in the present terminology, that his addressees remained unable to recognise and reduce deviations from 'qualifications', or alternatively prevent them, nor could they do this for 'quality sets'. The only (general) advice he eventually felt entitled to give was 'keep thinking', 'consider one's place'.

This lack of success does not make Fischhoff's approach irrelevant or un-interesting. It is rather his proposal to include in his 'first part' only one type of act, i.e. decision-making, that stands refuted. At least in retrospect it appears obvious that also the activities of the rapist himself would need to be included to transmit the qualifications of becoming neither rapist nor raped. This suggests extending the first part, and including additional activities.

One obvious extension would be to create 'knowledgeable collectives', that is collectives of as many activities as needed to create a 'first part' that would be independent of further forms of use (and re-establish the separation between first and second part). Such activities would have to interact, that is mutually 'advise' each other, or rather advise their actors to conform to qualifications that would make the collectives show desired properties (e.g. avoiding social violence). What needs to be known, therefore, is how the qualifications are to be transmitted without imposing a dictate [5].
Vahl [6] tried to identify and test advice to help workers in service teams self-organise and improve their functioning. First she translated existing interactions into statements that would transmit qualifications ('if observation X, do action A'). Next she taught workers how to change either the X, or A, or both X and A as three kinds of advice to achieve changes in the interactions within the teams-for example to reduce cost or increase service to clients. Workers had to react mutually, continuously, as soon as any worker implemented the advice. Eventually they did appear to create 'knowledgeable collectives'.

Comments

Although daily life nowadays is thoroughly permeated by the results of science, it still appears to consist mainly of advice. This is used to warn people about dangers, to support the needy, to teach the facts of life. It will be especially important to recognise 'good' advice, therefore. Interestingly, this is precisely what science is not good at. The main reason of course is that advice implies preferences. In other words, advisees may not accept advice, whatever its quality-in contrast to (correct) predictions of what shall happen.

A prime example of an area where recognising 'good' advice is important is the creation and management of organisations, businesses and other forms of service. Both activities usually are complex, as usually it is not easy to recognise what is relevant. The literature is replete, therefore, with (many types of) advice. One example is the work of [7], and of [8], another what is known as 'action research' [9, 10]. Much effort is spent on exploring how to improve the advice. There still is much uncertainty as to the results.

Another but related area is 'research' in general. At present there are many 'methodologies' to do research, or in the terminology of this paper 'quality sets'-containing advice to do certain things where and when. A major component is how observations are ordered and transmitted to constitute knowledge. This 'how' has been standardised as the 'first part' in advice (see above). There is an increasing interest to create something like 'knowledgeable collectives', however, using suitable forms of transmission [11].

The main reason seems an awareness that using standard research methods outside standard areas easily results in what is then diagnosed as dehumanising, objectifying, colonising, etc. Similar criticisms are levelled at 'methodologies' for management [12, 13, 14]. They also tend to force people into a mould. The work of Checkland on giving advice to advice managers provides an example. In relation to 'testing' he appears interested mainly in whether advice leads to desirable results, not in whether 'quality' can be maintained.

Conclusion

The question that has been central to this paper was how and to what extent advice can be tested. 'Advice' was taken to consist of 'advisable' actions that advisees may implement, but not necessarily. 'Testing' was assumed to mean providing help to decide whether or not to accept advice. It was proposed to do so on two levels: to recognise whether some (new) advice belongs to a 'tested' set of advice, and to maintain the qualifications transmitted by such advice, if recognised.

Advice permeates our life. Examples from three areas were considered. The first is the reduction of social violence, in particular by those in the role of the victim. The second is the improvement of organisations, and how they are managed. The third is the improvement of research, to support the accumulation of knowledge. In each area some form of 'testing' seems possible a priori, or in other words, using a design other than an a posteriori evaluation.

References

1. Fischhoff, B. (1992), Giving advice. Decision theory perspectives on sexual assault. American Psychologist 47/4/577-588.
2. Suppes, P. and Zinnes, J.L. (1963), Basic measurement theory. In: Luce, R.D, Bush, R.R. and Galanter, E. (1963), Handbook of mathematical psychology. Wiley, New York
3. Tufte, E.R. (1997), Visual explanations. Graphics Press, Cheshire, Conn.4. Zeeuw, G. de (1996), Does main-stream system theory support innovation? In: Rebernik, M., and Mulej, M., eds., Proceedings of the 3rd International Conference on Linking Systems Thinking, Innovation, Quality and Entrepreneurship (STIQUE '96), pp. 333-343
5. Bradburne, J. (2000), Interaction in the museum. Observing supporting learning. Ph.D. University of Amsterdam (ch. 2 and 3)
6. Vahl M (1994), Improving mental health services in Calderdale. Centre for Systems Studies, University of Hull, Hull
7. Checkland, P. (1981), Systems thinking, Systems practice. Wiley, Chichester
8. Warfield, J.N. (1990), A science of generic design. Managing complexity through systems design. Intersystems Publications, Salinas
9. Denzin, N.K. and Lincoln, Y.S., Eds. (1994), Handbook of Qualitative Research. Sage, Thousand Oaks, CA.
10. Reason, P., Ed., (1988), Human Inquiry in Action: Developments in New Paradigm Research, Sage, London.
11. Zeeuw, G. de (1997), Second order organizational research. In: J. Achterbergh, R. Espejo, H. Regtering, M. Schwaninger (eds), Organizational Cybernetics. Nijmegen Business School, Nijmegen
12. Zeeuw, G. de (2000), On constructivism: complete collectives and the evolution of research. To be published in Foundations of Science.
13. Pettigrew, A.M. (1997), Context and acion in the transformation of the Firm. In: Pugh, D.S. ( ed.), Organization theory, selected readings. Penguin books, London
14. Semler, R. (1993), Managing without managers: an unorthodox approach to coping with complexity. In: Espejo, R. and Schwaninger, M. (eds), Organisational fitness, corporate effectiveness through management cybernetics. Campus, New York
 return