Preferences and research

1. Introduction

It is bad practice to adopt beliefs on the ground that they make us feel comfortable. We should try to find out how things are, even if they are not as we would like them to be. We should try to fit the contents of our minds to the state of the world - the mind-to-world direction of fit. It is otherwise with our desires. If the world is not as we would like, we should try to change the world to match the demands of our minds - the world-to-mind direction of fit.

We shall consider whether there is scope for something in between. We seek beliefs, in the formation of which preferences as to how the world should be may have a legitimate role to play. (It seems plain that the other things in between, desires in the adoption of which beliefs may have a legitimate role to play, are widespread. It is a good idea to be well informed about the likely consequences either of having or of fulfilling various desires.)

When we speak of preferences, we shall mean views that are not themselves conclusions reached legitimately in the course of study of the world. Existing theories about the world may guide current study, and such guidance is likely to be entirely legitimate so long as there is no dogmatic view that the existing theories must never be abandoned. We are instead concerned with preferences for how the world should be that are not themselves grounded in research. For example, a physicist might want the world to conform to particularly simple equations, or a historian might want his or her country to look glorious, or alternatively to blame it for past evils. The physicist, or the historian, might then collect evidence and draw conclusions accordingly. That would be a very dubious way to work.

2. Mathematics and the natural sciences

In mathematics and the natural sciences, there is assuredly no place in the formation of beliefs for preferences that are specific to the current research project. Preferences may lead us to choose areas of work, but once we set to work we must put up with whatever conclusions deduction and experiment force upon us. We may indeed take it as a mark of scientific respectability that preferences that are specific to current research projects have no role in the formation of beliefs.

Having said that, not all preferences are inappropriate. There may be a preference to work in ways that are regarded as respectable. Such ways might range from the most general, such as always testing a hypothesis on data other than those which suggested it, to ways specific to particular disciplines, such as checking the purity of any supplies of chemicals before using them. A preference to use ways of working that were regarded as respectable might adversely affect research if the ways were not really the best ones, but on the whole the preference would be a worthy one. There would also be some protection derived from the fact that general ways of working are very widely used, so if they were inappropriate on any more than a trivial proportion of occasions there would be a good chance that someone would have noticed and would have reported the problem. This would however not cover the risk that ways of working, while good in general, were for some reason not well suited to a particular research project.

3. The social sciences

It would seem that the same exclusion of preferences from belief formation should apply to the social sciences, disciplines that aim to understand the ways in which human beings behave. It may however be harder to keep preferences at bay. Few of us would like to see molecules behave in a particular way. Far more of us would like to see people behave in a particular way.

3.1 Hoping for certain results

We might be concerned that preferences would have a direct influence, with conclusions as to how people in a given society currently behaved being skewed simply in order to end up with a satisfying picture.

That would however be so plainly unjustified that it should be ruled out from the start by any honest researcher. And if it happened unconsciously, one would hope to hear harsh criticism from others who commented on the work.

This is not to say that such skewing would never happen. Our claim here is only that the mere desire for comforting conclusions would be seen not to suffice to outweigh the arguments against its having any influence. But another motive, to put results to practical use, might appear to justify a role for preferences even if it did not really do so. And in any case, a role for preferences might be found to be inevitable.

3.2 Using results

3.2.1 Desired changes to societies

A difference between the natural sciences and at least some of the social sciences emerges when we consider the use of results.

In the natural sciences, it is reasonably easy to maintain a clear boundary between getting results and using them. We find out how atoms or molecules or medications or large pieces of steel behave under certain conditions, then we decide what we want to do. In choosing what to do and then doing it, we use the knowledge obtained.

In the social sciences, it is harder to maintain a clear boundary. To say that a given society functions in a certain way under current conditions can easily be to make a moral comment on whether it functions as it should. That can in turn create pressure to act so as to change the society.

It is not necessary that such a train of thought should be followed. If one studies societies in general, or a large class of societies that happens to include one's own, one may easily remain detached from practical commitment. But if one concentrates on a society within which one has the capacity to act, practical commitment becomes more likely.

When certain conclusions would lead on to such commitments, preferences may come to influence the conclusions reached. One may have existing views on how society should be, and then work in ways which would make it more likely that one would reach conclusions that would encourage people to move society in that direction. If for example one had an existing view that significant economic inequality was deplorable, one might study its effects in ways that highlighted its adverse consequences.

The mechanism would not be to set a course directly to the desired conclusions. Instead it would be to choose approaches and concepts to use. Some approaches and concepts would be ruled out by their tendency to yield unsatisfactory results. But options would still remain. And the choice might well be influenced by one's preference to work in ways that would lead to certain types of conclusion. One could re-do the work using other approaches and concepts and compare the results, but it is unlikely that one would have the time to do so, particularly when different data would need to be collected under different approaches. Even if one did so, there would still be scope to regard results obtained using one's favoured approach and concepts as intellectually more respectable than results obtained in other ways. The social sciences do not tend to produce results that can be subjected to decisive experiment, so there would be scope to use judgement as to which were the better results.

One could still say that this should not be so, and that the correct method would be first to reach conclusions without any influence from preferences and then to choose any actions to take, based on a combination of those conclusions and one's preferences. In that way, preferences could be respected without their interfering with the study of a society. But there is a clear risk that such a noble course would not be followed.

3.2.2 The general practical emphasis

Not only preferences for particular changes, but also a general emphasis on the practical use of research, may open the door for preferences to influence the approaches and concepts used, and hence the conclusions reached.

Particular pieces of research in the natural sciences may be started for practical reasons, for example to discover how to build better aircraft or how to cure certain diseases. But there is still a sense of seeking some pure truth that would be worth pursuing for its own sake. In the social sciences, the primary motive will often be to solve some practical problem, and it is less obvious that there is an ideal of pure truth to inspire work. We might well work in the natural sciences without any practical goal in view. It is much less clear that we would bother with the social sciences if there were no social concerns to address. We might for example decide not to bother because there would be little prospect of reaching results that approached the robustness even of results in the biological sciences. That lack of robustness would put in question the very concept of a pure truth to which one might aspire.

Now suppose that work is indeed motivated by a desire to solve practical problems. Then conclusions will only be of much interest if there is a good prospect that they will help to solve those problems. Here we come up against the fact that researchers and the objects of study are all human beings, with priorities, values and attitudes. The priorities, values and attitudes of the objects of study must be respected in the formulation and implementation of any solution if it is to succeed. The easiest way to ensure such respect is for researchers to internalise those priorities, values and attitudes, so that conclusions reached are more likely to be usable in devising solutions that will work. Full internalisation might be out of reach, but at least a degree of empathy with the objects of study would be very helpful. And that would in turn be likely to sway the selection of approaches and concepts.

There would be another way to look at this, which might eliminate the need for empathy. It would be to say that researchers should remain detached, but that they should take the priorities, values and attitudes of objects of study to be features of them which entered into the definition of the problems to be solved. There would then be scope to reach conclusions like "For a society of people with such-and-such priorities, values and attitudes, such-and-such policies would be likely to produce such-and-such results". There would however be no guarantee that this alternative way would be chosen by researchers. And it might not be feasible to define the priorities, values and attitudes in a detached way, without having one's thoughts guided by the concepts that were used by the objects of study.

3.3 Might a role for preferences be inevitable?

Establishing that preferences that were specific to current research projects sometimes played a role in the formation of beliefs would not show that it was legitimate for them to play such a role. And it would be hard to argue that it was legitimate.

Having said that, preferences might have to play a role. Then we might still think the role illegitimate but would have to put up with it.

A role for preferences would be inevitable if that was the only way to choose between the approaches and concepts that remained available once the approaches and concepts that were known to yield poor results had been excluded.

It might be the only way because alternative approaches and concepts could not be ranked entirely in a detached scientific way. Rather, they would be ranked at least partly by reference to whether they seemed to focus on the factors that researchers considered important. The selection of such factors could easily vary with, for example, the political preferences of researchers.

It might also be that a given choice of approach and concepts could not be judged conclusively after the event, even if results obtained using other approaches and concepts were available for comparison. Such an inability to judge conclusively would follow from a lack of decisive tests of the quality of results that were independent of the approaches and concepts used in reaching those results.

If a role for preferences were inevitable and the results of choices made under their influence were not open to conclusive judgement, we would be faced with an influence of preferences that might in general be thought deplorable, but that could not be eliminated in particular cases. This would seem to degrade the relevant discipline as a whole, but without giving any practical recommendation for its improvement. We would then see a contrast with the natural sciences, in which it is a mark of good work that preferences which are specific to current research projects play no such role.

4. Politics

Someone in a position to influence government policy, whether as a member of the executive, of the legislature, of the media or even of the voting public, should perhaps work out the best thing to do in a dispassionate way, without giving priority to their own preferences. (Compare John Stuart Mill's view, in Considerations on Representative Government, chapter 10, section 1, that a citizen's vote "is strictly a matter of duty; he is bound to give it according to his best and most conscientious opinion of the public good".)

It is however very likely that when someone works out what he or she honestly believes to be the best choice for his or her country or municipality, he or she will in fact be influenced by his or her preferences more than by the preferences of other individuals or modest groups of individuals, and perhaps more than by any identifiable preference of the majority of the relevant population. And if he or she considers his or her own preferences to arise not out of considerations of personal benefit but out of a proper understanding of the public good, this is even more likely to be so because it will seem that the competing preferences of others are based on factual mistakes.

Would this be an example of a research project, the project of working out what to do, in which the answers obtained were influenced by preferences? And if so, would the influence be legitimate?

Our first question is, would there be any research project here? It would seem that there would be. There might not be a single correct answer as to which policy to adopt, but there would be some options that were plausible under most value systems, some that were only plausible under a fairly restricted range of value systems, and some that were implausible under nearly all value systems. So there would at least be the project of identifying relevant value systems and testing the plausibility of options.

Here, preferences could exert an influence. They would not lead directly to the identification of the better policies. Instead they would lead to the identification of value systems worth considering, or of value systems under which any policy would have to be plausible in order to be acceptable. Then policies could be measured up against appropriate value systems.

However, once people got past the stage of sorting policy options into those which were and those which were not acceptable by reference to appropriate value systems, and had to select specific policies, the process could no longer sensibly be called a research project. It would lack independent ways to judge whether choices were correct. Then it would be time to stop worrying about the roles of preferences in the way that we have been worrying, although we might have other worries about their roles.

The involvement of preferences in identifying value systems to use when sorting policy options into those to consider and those to exclude would appear to be intellectually respectable. But there would be an important constraint. Respectability would only be maintained if the value systems which were used to sort options had not been preferred so as to ensure that certain policy options remained in play while others were excluded. They should be value systems that would be preferred independently of the immediate policy question.

Respecting this constraint might not be easy. We tend to judge value systems partly by how they would apply in specific situations. And it would be possible for a value system that had hitherto led to perfectly acceptable conclusions as to what to do in specific situations suddenly to appear unacceptable, either because it would endorse what was intuitively an unacceptable option in the current situation or because it would lead to the rejection of what was intuitively a very appealing option.

Should we really exclude current situations from the resources we have for deciding between value systems, even when the message from a current situation about the merits of a value system is a very strong one? Or should we be tempted by the view set out by Robert Nozick (Philosophical Explanations, chapter 4, section 1, subheading Nonrandom Weighting)? He sets out how we assign weights to the various considerations that might bear on a decision. The assignment of weights on one occasion will influence assignments on future occasions by setting a precedent, but not an inviolable one. He also argues that the assignment of weights on a given occasion can set a precedent under which that assignment is itself subsumed. The relevance of this approach to our discussion is that it offers a way in which consideration of a current situation could legitimately influence a choice of values under which that case would be judged. We here leave open the question of whether it would be wise for us to make use of Nozick's proposal.

5. Legal judgements

5.1 The lack of certainty

There are legal cases in which, even when all the facts are known and agreed, it is unclear what the outcome of litigation will be. We can see this as cases progress up through hierarchies of courts, and higher courts sometimes reverse the decisions of lower courts on points of law without any new findings of fact.

The most general characterisation of what is involved here would be that the outcome of applying the law to the facts was not obvious. We might see a research project of finding out how the law should apply to the facts, and then ask whether preferences had any legitimate role to play in that research. We might rule out in advance preferences as to the outcome of the instant case, but still find a legitimate role for preferences as to how to apply the law so as to produce intuitively satisfactory outcomes more generally.

We should not however stop there. It is worth thinking about how we might conceptualise the uncertainty of outcome.

5.2 Conceptualisations

5.2.1 The law as an incomplete algorithm

One conceptualisation would be that the law was an incomplete algorithm for computing decisions from sets of facts. It would be impossible for legislators to think of all possible sets of facts in advance. And the use of natural languages to write both laws and accounts of facts would make it hard even to define boundaries with sufficient precision to ensure that every case would be covered, for example by defining cases to which one rule applied and then saying that some other rule applied to all other cases.

Under this conceptualisation, courts would be seen as filling in the gaps. They would create law for instant cases and, if there were a system of binding precedent, for future cases. Then one could see a legitimate role for preferences at the level of deciding instant cases, but on the other hand doing so would hardly count as a research project. One could however identify a research project of working out the range of ways in which the law could legitimately be developed. And at that level there might be a role for preferences as to the standards of legitimacy to impose, comparable to the role for preferences in identifying relevant value systems that we mentioned in connection with politics in section 4.

5.2.2 Correct answers waiting to be found

A second conceptualisation would be that the correct answer on the basis of the instant facts was already given by the law, but judges sometimes could not see it. It would be implausible to think that the law already held answers written in invisible ink. But it would be perfectly plausible to think that what appeared to be courts creating the law by filling in gaps was in fact their attempt, sometimes successful and sometimes not, to discover the correct way to fill in the gaps, where there was in fact a single correct way.

(We might draw an analogy with advances in mathematics, where the correct new results are implied by the existing body of mathematics but one has to work hard to tease out the implications of that body. The analogy would however suffer from the defects that we cannot expect rigorous proof in questions of law, and that we should not take it for granted that there are single correct ways forward, as we usually although not always can take it for granted in mathematics.)

An obvious point of reference is Judge Hercules, whom Ronald Dworkin invented in chapter 7 of Law's Empire. Hercules studies the law to date, including all the previous decided cases, and works out how to decide a new case in a way that is consistent with all of the law to date. He has superhuman powers of reasoning to take him to the right answer. He works like a novelist who is writing and publishing a work chapter by chapter, and who must make each new chapter, each new development of a character, and so on, consistent with what has already been published.

This point of reference indicates, but does not quite capture, what we have in mind as our second conceptualisation. Dworkin fills in a specific constraint on Hercules' thought, the need for consistency with the existing legal tradition. This is right in order to take Dworkin's programme forward, but we would like to keep things more open and simply consider whether it might be possible to work out a single correct way to fill in what appeared to be a gap in the law, whether by seeking consistency with the tradition or in some other way (although consistency or a good justification for inconsistency would be expected anyway).

Now suppose that this second conceptualisation were used. Identification of how to fill in a gap in the law would then look like a research project. What legitimate role might there be for preferences?

One possibility would be a preference for working in a sensible way to establish how a gap should be filled in. This would be analogous to a harmless preference for sound scientific methods.

In principle, there should not be other legitimate roles for preferences. The premise of the conceptualisation is that there is a correct answer to be found. This makes it analogous to a scientific enterprise. The same consideration would rule out the idea that there might be scope to work in different ways that might lead to conflicting answers.

Having said that, there would be the difficulty that since human judges are not as capable as Hercules, even judges working in the same way might reach different conclusions. And even if they all agreed, it would not be possible to be sure that the correct answer had been found.

5.2.3 Answers automatically becoming correct

One might say that an answer automatically became the correct answer by virtue of its having been chosen by a suitably elevated court, at least in a legal system in which decisions set precedents. (There would be an imprecise echo of Robert Nozick's idea of setting a precedent under which the decision was itself subsumed.)

If one did say that, would there still be a research project? There would be, in the sense that the court would have to work out the options for filling in the gap in the law. Here one could see an application of the view that Dworkin took. Some options would be available because they would represent coherent developments of the law, while others would be ruled out because they would not fit with the existing legal tradition. And there would be a role for a general preference to do this job well by applying general standards of coherence that would apply in non-legal matters too. (That is, the law would not have its own special standard of coherence, although it would have ways in which incoherence could arise that were specific to legal matters.) But there would not seem to be a legitimate role for other preferences that had their standing independently of the relevant legal tradition, such as ethical preferences. What passed the test of coherence would be determined by standards which were independent of any such preferences.

After the options had been listed, one of them would have to be chosen. Then there might well be a role for ethical preferences, either in relation to the instant case or in relation to the foreseen consequences of setting one precedent rather than another. But at that stage there would no longer be a research project. We take it to be essential to a research project that conclusions should be constrained by considerations outside the researcher's psyche. We are exploring ways in which we might move a little way away from being entirely constrained by the world, so as to find some role for preferences. But in a research project, preferences need to play at most only a modest role. If there is scope for them to dominate, the work is no longer a research project.

6. Analytic philosophy

Analytic philosophy is a curious discipline. Questions are posed and correct answers to them are sought. But disagreement is widespread, and there is nothing analogous to decisive scientific experiments that could be used to rule out enough of the proposed answers to bring philosophers close to resolving disputes. (There is experimental philosophy, but it is primarily concerned with elucidating the views of a wider range of people than philosophy professors so as to draw conclusions about what is or is not intuitively clear.)

So does analytic work on a philosophical question constitute a research project? It would seem that it does. There is certainly a sense that there should be an answer out there in the world, to be discovered and then defended against objections, rather than a sense that one is only formulating views that one is inclined to express. And some possible answers do get ruled out by some philosophers, although some of those excluded answers may still be supported by other philosophers.

We should also consider questions where the conflict is not so direct. Rather than some philosophers believing that an answer is correct and others believing that it is mistaken, we may find some believing that it is adequate, or that the approach is appropriate, while others believe that the answer is inadequate, or that the approach is inappropriate.

The analysis of knowledge is a good example. The traditional analysis as justified true belief is now generally regarded as inadequate. But some epistemologists seek to modify it, adding additional conditions to rescue it from Gettier-like objections, while others say this is not the right way to study knowledge. They may want to take epistemology in a new direction such as virtue epistemology, or they may want to call a halt to the programme of analysis by taking knowledge as fundamental, as in Timothy Williamson's knowledge first approach.

Claims that some existing answer within a given approach is inadequate and needs to be supplemented present no challenge to the idea that a research project is being pursued. Here there would not appear to be a legitimate role for preferences as to what the answer should be. The only legitimate preferences would be for clear thought, the careful consideration of objections, and so on.

A choice between approaches might be a research project, but only after the approaches had been tried by oneself or by others. Approaches would be used in what would be research projects in their own right. Then the project of selecting a favoured approach would be taken forward by considering which approach had shown itself to be the most satisfactory.

In such a project of choosing between approaches, preferences might have a role. Approaches might be more or less satisfactory in several ways, for example the range of the issues they covered, the smallness of the range of areas of difficulty, the degree to which the conclusions to which they led appeared to be robust in the face of objections, and their goodness of fit with approaches favoured in other areas of philosophy. It would be perfectly possible for one approach to do particularly well in some such ways while another did particularly well in other such ways. Then the selection of an approach would require a sense of the relative importance of the different ways in which some approaches might be better or worse than others, and assignments of relative importance could reflect preferences.

We seem to have here a reprise of the position one may find in the social sciences when preferences may be necessary in order to complete a project of identifying the best approach, but the influence of preferences may still be deplored.

One might prefer to abandon the project of choosing an approach and instead keep all approaches in play, providing a toolbox from which one could take first one approach, then another. And given that the task of data collection is much less onerous in philosophy than in the social sciences, one could easily use several different approaches in parallel.

If one did use several approaches in parallel, one might be faced with contradictory answers to the same philosophical question. One might for example find that under one approach a subject in a given situation would be credited with knowledge, while under another approach he or she would not be credited with knowledge.

A possible response would be to fragment some relevant concept, such as the concept of knowledge. A subject might be said to have knowledge-under-justified-true-belief-like-approaches but not to have knowledge-under-virtue-epistemology.

That would however damage the coherence of an area of study. It is the idea of a single conception of knowledge that gives epistemology its shape and character, and similarly for other areas of philosophy. The dissolution of a problem is not always the ideal solution.