Robert Dingwall

Institute for the Study of Genetics, Biorisks and Society

University of Nottingham

Today I am going to concentrate on the distinctive themes of the sociological approach, although in practice there is a pretty soft boundary between social psychology and sociology - there are no real fundamental differences between the way Maria and I would approach a particular problem. The question is how do sociologists think about error?

Yesterday Maria was saying errors are not the result of inevitable human failures: people make mistakes at work all the time. Errors are social phenomena but they only become consequential if they occur within a social system that doesn't identify and remedy them before any damage occurs. This then poses a set of issues about strategies that are oriented to individual performance because, if errors result from system rather than individual failures, they can only be marginally influenced by individualist strategies. To improve performance we need to know rather a lot about what constitutes adequate performance - which is very hard to define in professional practice especially because people are so intrinsically variable. It is the fundamental problem of evidence-based medicine, of any sort of rule-based solution to issues in professional practice, that you cannot reduce that practice to algorithms because, without space for professional judgement, you end up following rules and killing the patient. That is a fundamental weakness in the drive to bureaucratise professional practice. Rule-based approaches on the whole end up imposing an average level of mediocrity: good performance requires flexibility but this also allows the possibility of bad performance. We can see in the universities where the assessment of teaching quality is driving out teaching quality because everyone is teaching to the script. Improvisation and innovation that might excite students are driven out by the need to plan in advance so that courses are turned into a kind of moral contract and then delivered according to that contract. If you walk in one morning and go 'Gee, I read this really interesting paper which has inspired me and I want it to inspire you' you can get the reply, 'it wasn't in the handout at the beginning of the course, it is not on the reading list, so why are you doing this?'

If we are approaching error as a systemic phenomenon, we need to answer the questions 'Do we have reliable and valid knowledge of the system? Does introducing extra complexity into that system actually improve its performance? One useful approach (that you hinted at Betsy) is that it can be helpful to try to model what an error free organisation would look like. It is a bit like the way we can use the idea of a perfectly free market in economics. I am sure that Martin will tell us that the perfect market is a useful idealisation that helps us to notice things about the real world but we don't expect to find one because of the practical difficulty in producing it. An error free organisation channels information relevant to the processes and outcomes in real time. We have to ensure that that information is unambiguous which runs into some major problems in the philosophy of language - the impossibility of writing an exhaustive specification of dealing with anything - what the sociologist Harold Garfinkel called the etcetera clause because any statement depends on its context for meaning: the rules for noughts and crosses do not teach someone how to play the game because they depend on a whole bunch of assumptions and cultural practices in which they are embedded - like you do not use an eraser to erase your opponent's marks. There is nothing in the rules about that, it is just assumed that you do not do that sort of thing. You would have to ensure that information was never overlooked. You would have to be systematically dispelling ignorance so that that everything is perfectly clear and intelligible with continuous real time feedback with the continuous possibility of real time realignments of actions to fit environments. By posing the problem that way, one can ask what is it that stops the world working that way. Continuing the analogy of the perfect market, many of the most interesting things that economists have to say are about what stops the real world from being perfect: for example, if information is not available at appropriate times in a usable form, one of the reasons is the sheer cost of doing this in relation to the resultant benefits. The costs of eliminating patient death and harm caused by the NHS would be wholly disproportionate to the benefits, which is not to say that one should not strive to minimise that level but that resources are finite because healthcare is in competition with other social goods. So the question becomes, 'what is the minimum amount of error, injury and serious harm that is compatible with that amount of investment without diverting resources from other activities of benefit to patients?'

Sociologists have classically identified several types of information failure.

· Things we just don't know about - can be addressed by improved search strategies. · Things we know about but not are not fully appreciated - these cause a false sense of security - we think we know something but actually we don't. Examples are: Ø Pressure from competing tasks Ø Distrust of information sources - the example Nick gave yesterday 'Are they doing it at Barts? If they are not doing it at Barts but it is being done at Guys but everyone knows they are a bunch of charlatans and we shouldn't take any notice… Ø Decoyed by different problem - exemplified by Diane Vaughan's work on the Challenger shuttle launch disaster. Everyone knew that the O-rings were a problem but they were one of many problems that NASA knew they ought to fix 'at some time' rather than among the dozen problems that were of current concern but in the event did not cause a catastrophic failure. Ø The problem of sorting out relevant from irrelevant information - again one of the things that sociologists do contribute is a radical scepticism about whether or not more information is a good thing or just causes more confusion. Ø Known but not fully assembled - one of the things that Tony touched upon was 'social cognition'. Sociologists and psychologists are increasingly seeing cognition as embedded in networks and organisations where information is distributed between different actors so the whole picture is invisible to any of them - people have different parts of the picture, the challenge is to join the parts together to make a whole but also in recognising that knowledge is not owned by any one of them. Knowledge is in the relationships between them. There is very interesting work on aircraft and ship navigation from this direction. Ø Things that are available to be known but don't fit current frames - part of the response to information overload is the creation of bounded information zones. This is part of daily life, 'how do we manage to avoid, dissattend or ignore most of the stimuli that impact on us?' The division of labour and the nature of modern organisations creates these zones to enable us to decide what we need to know to get through the daily workload information may well come in (as it did about the extent of the O-rings) that doesn't fit the current frame which is 'we are worried about other things and we have these accumulated reports of O-ring erosion but we are not really focussed on this problem': they do not fit the frame and we can put them to one side.

Tony has introduced some of the ideas that go together with what sociologists of science and technology call actor network theory, which has some helpful ideas about the way in which we think of networks and actors within them. This, I think, is one of the defining differences between sociology and psychology because this theory is not interested in individual minds or states: actors are points in the networks and it is the system that has properties and defines what an actor is, the possibilities of action, and the structural and cultural framework that constitutes the actor and the action. The theory is interesting in its exploration of material objects as actors. I am passing around a material object that is also an actor. This £5 note encodes a lot of information in any commercial transaction. The great Austrian economist, Hayek, wrote a wonderful paper about money as encoded information and points out that a currency note brings with it a whole history, that we need know nothing about how this note comes to represent £5 but it can still be an actor that communicates information between people in a social system. When technology is embedded in social and cultural systems, we need to understand the interface between people and technology. I once heard a wonderful paper about operators in French nuclear power plants. There are two ways in which the nuclear reactors can be controlled: either operate one control that moves the graphite rods up and down or use another control that adds bits of boron, which damps down the reactor. The operators try to run the reactor as near the safety limit as possible to maximise power output. Although the graphite rods are the more sensitive control. the operators just use the boron control which requires the reactor to be run into the critical zone before boron can be added to just bring the reactor to within the safety level. So the alarm bells are ringing and the operators then carefully titrate the dose to damp the pile - which is a lot more fun for the operators. These guys were running the plant in the same way as in French cookery - a dash of this, a dash of that, no precise recipes. This is a nice example of the safety features being embedded in technology but people using the system differently.

To conclude, the way we tend to think in a sociological direction is that all systems will fail in different ways, Complexity adds to the ways of possible failure - you solve one problem but probably introduce another. I am not sure I really believe in error homeostasis but systems always have an optimum error rate: you know, how do engineers build bridges? They push the technology past the limit until a bridge falls down so they know they have gone over the limit. A system that is not making errors and failing from time to time, it is probably not innovating or progressing. The question is how do we manage those errors and get them to an acceptable level?

Questions BF A quick suggestion about the Challenger disaster, there is another great source the 'Visual Display of Information' that talks about how if the engineers who had done the analysis had looked at the graphs differently they could have spotted the temperature differences. This is a great reference book.

MP You raised the point about how much investment to eliminate errors in the NHS. My view is that we are so far away from anything like elimination that the question is purely theoretical. If I give you an example, about ten years ago, I was asked to investigate several instances of poor service in a leading maternity hospital. Mothers who complained were interviewed and what was revealed from the interviews was just a bloody awful service. It didn't take much investment to make a marked improvement. As Anne said, things are being done so badly that we are light years away from being able to ask the question. Should we invest more to reduce the error rate?

RD I think there is some justice in that. On the other hand, it is a question that we should be asking and, as has been mentioned several times, do we end up spending a lot of money eliminating relatively cheap inconsequential errors simply because we are focussed on the error rate or should we be more selective when investing in the much smaller proportion of seriously consequential errors which may be much harder to deal with. In a previous part of my career, I have done a lot of stuff on divorce mediation, which had a huge push behind it. Trouble was, it was pitched as 'this is the way we solve the 10% of divorces that are really seriously bitter' and it didn't do that - what it did was to increase the level of intervention in the 90% that were going to be reasonably amicable, as far as divorce can be, that so mediation increased costs and delays but did not address the system problem but simply problematised reasonably non-problematic divorces. So we should continue to ask this question.

OH I think complexity is a very important source of errors in healthcare and it is combined with and related to change so we can only have a partial knowledge of newly introduced technology. We cannot predict the side effects of a drug and the greater the complexity the greater the chance of more things interacting. The problem in healthcare is the pace of change and the complexity and thus errors will grow just as rapidly, so you can say that things can only get worse.

TC This may mean that addressing errors gets more important too.