Contact

The prospect of genetically modified crops and foods has become a political hot potato in Britain. Food industry executives, government advisors and biotechnologists have been caught unaware of the strength and persistence of public concern. The pilot study by SPRU on multi-criteria mapping of genetically modified crop in agricultural systems in the UK shows how in such an overheated political arena highly polarized disputants can participate constructively in discussion and regulatory appraisal and create a " map " of the debate surrounding GM crops. The social appraisal of technological risk has always been a difficult undertaking. With diminishing public confidence in traditional expert institutions and procedures, the difficulties seem to be growing. In some quarters, there has been a tendency to see this as mere cultural fashion – reflecting the ephemeral anxieties of the " risk society " (Giddens 1991; Beck 1992; Lash et al. 1996). Under this view, the answer lies in basing the regulatory appraisal process on " sound science " , keeping it as free as possible from the messy and emotive business of politics. In this regard, aspiring " scientific " techniques, such as probabilistic risk assessment and cost-benefit analysis, are sometimes held to offer rational, robust and objective aids to " science based " decision-making. Unfortunately, the difficulties in the social appraisal of technological risk are not so easily resolved. Although the language of " sound science " remains prominent, the available methods and procedures manifestly fail to deliver on the rhetoric. This is not simply a matter of practice failing to live up to potential. Ironically , the reasons for the failure of " sound sci

The social appraisal of technological risk has always been a difficult undertaking.With diminishing public confidence in traditional expert institutions and procedures, the difficulties seem to be growing.In some quarters, there has been a tendency to see this as mere cultural fashion -reflecting the ephemeral anxieties of the "risk society" (Giddens 1991;Beck 1992;Lash et al. 1996).Under this view, the answer lies in basing the regulatory appraisal process on "sound science", keeping it as free as possible from the messy and emotive business of politics.In this regard, aspiring "scientific" techniques, such as probabilistic risk assessment and cost-benefit analysis, are sometimes held to offer rational, robust and objective aids to "science based" decision-making.
Unfortunately, the difficulties in the social appraisal of technological risk are not so easily resolved.Although the language of "sound science" remains prominent, the available methods and procedures manifestly fail to deliver on the rhetoric.This is not simply a matter of practice failing to live up to potential.Ironically, the reasons for the failure of "sound sci-ence" as a basis for policy making are fundamental to the science of probability, risk and rational choice itself.
First, there's the problem of robustness.The scientific assessment of "risk" -defined in the formal sense shown in Figure 1 below -requires at least some basis for assigning probabilities (Neumann, Morgenstern 1949;Juce and Raiffa 1957).Under conditions of "uncertainty" and "ambiguity", however, the novelty, idiosyncrasy or complexity of the processes involved can preclude this (Knight 1921;  1921).Under even less tractable conditions of "ignorance", not only the probabilities, but also some possibilities may be unknown (Loasby 1976;Smithson 1989;Wynne 1992;Stirling 1998).Even the most authoritative of experts don't know what they don't know!Here, as arguably already experienced with BSE, the ozone hole and endocrinedisrupting chemicals -and potentially with genetic modification technologies -we face the prospect of surprise.The conditions of "ambiguity", "uncertainty" and "ignorance" are fundamental to the scientific characterisation of risk.Yet, by definition, the techniques of probabilistic risk assessment fail to address these aspects of incertitude.Such techniques are therefore inapplicable to some important and pervasive aspects of environmental decision making.Far from upholding science, persistent adherence to inapplicable methods is quite simply unscientific.
Second, there's the issue of objectivity.Although often very precisely formulated, the results obtained by risk assessment can be highly sensitive to starting assumptions.In areas ranging from energy, through transport to chemicals, the conclusions of different risk assessments often vary radically -sometimes by many orders of magnitude (Stirling 1997a).This can have profound implications for conclusions over the relative performance of different technology or policy options.Yet the choice of which assumptions to adopt are often contingent and intrinsically subjective (Royal Commission on Environmental Pollution 1998;National Research Council Committee on Risk Characterisation 1996).It is a fundamental finding in the mathematics of rational choice, that there can be no single objective resolution of divergent subjective perspectives (Bonner 1986;Bezembinder 1989).Indeed, the economist Ken Arrow was awarded a Nobel Prize in part for work that shows there can be no single definitive way to order a range of choices in a plural society (Arrow 1963;Kelly 1978).Therefore, instead of making apparently authoritative assertions over risk, a truly scientific approach lies in systematically exploring how the picture changes depending on how it is viewed.
These difficulties, among others, are recognised in the increasing international policy profile of the "precautionary principle" in the management of technological risk.Although the details are complex and hotly contested, the overarching "precautionary approach" builds on a series of straightforward and wellestablished ideas: that "prevention is better than cure"; that "the polluter should pay"; that we should look for "no regrets" options; that alternatives should be appraised at the level of production systems taken as a whole; and that we should recognise the intrinsic value of nonhuman -as well as human -life.More specifically, a precautionary approach acknowledges the complexity and variability of the real world and embodies a certain humility about scientific procedures and knowledge.It implies recognition of the vulnerability of the natural environment and prioritises the rights of those who stand to be affected by an activity, rather than those who stand to benefit from it.It requires scrutiny of all available alternatives and an examination of justifications and benefits as well as risks and costs.In short, a precautionary approach involves the adoption of longterm, holistic and inclusive perspectives in environmental protection (O'Riordan, Cameron 1994;Fisher, Harding 1999;Raffensberger, Tickner 1999).This is all very well, but to some it seems rather abstract and ambiguous -threatening to undermine science by opening the floodgates to subjectivity (Gray, Bewers 1998).Of course, it is true that -though not a panacea -the rigours of science remain essential to the process of risk management.The crucial question is, how can we combine the disciplines of science and precaution in the practical business of technology appraisal?Based on work underlying the recent European Commission Communication on Precaution (European Commission 2000; Stirling 1999), it is possible to construct a series of eight evaluative criteria against which the regulatory appraisal of risk can be assessed both in terms of its scientific rigour and its precautionary qualities.In short, these are as follows: • Humility: Does appraisal fully acknowledge the many sources of uncertainty, ignorance and subjectivity in appraisal?Does it avoid claims to complete or otherwise definitive knowledge?These are the considerations that have informed the development of the "multi-criteria mapping" technique employed in the present case study (Stirling 1997b).In short, the motivation behind this approach is to seek to combine the openness and qualitative flexibility of participatory deliberation with the clarity and focus of quantitative assessment.The specific case study with which this method has been piloted concerns the hotly contested debate over the use of genetically modified crops in UK agriculture.

The Debate over the Regulation of Genetically Modified (GM) Crops
The agricultural use of GM technologies is held by many to promise great benefits.On the other hand, others fear that the use of these technologies in agriculture may result in serious, irreversible harm.In the UK, formal regulatory appraisal of GM crops has centred on the question of whether or not they are "safe" for the environment and for human consumption.However, there is considerable scientific incertitude over the form and magnitudes of the possible effects and, as yet (by contrast with chemical or nuclear risks), little accumulated practical experience to draw upon.This has led to the evolution of a set of controls which are intended to be precautionary in nature -where it is accepted that action to avoid harm may be taken in the absence of scientific proof -with the conduct of risk assessment being required before experimental or commercial use of a particular genetically modified organism is allowed.
Despite this somewhat precautionary approach to risk regulation enshrined in the European Commission's Deliberate Release Directive (European Commission 1990) and the Novel Foods Regulation, the regulatory appraisal process has failed to gain confidence, either of Non-Governmental Organisations (NGOs), private industry (Mayer et al. 1996) or the general public (EPCAG 1997;Grove-White et al. 1997).This lack of confidence arises among other reasons: because the scope of the regulatory appraisal is still held in many quarters to be too narrow; because there is a general lack of trust in official reassurances of safety (particularly in the wake of BSE); and because justifications and benefits are not explicitly included in the evaluation process.Industry and regulators have expressed frustration, believing that the precautionary approach is being invoked in too burdensome a fashion, with unrealistic demands being made concerning absolute proof of safety.
It has also been almost impossible to gain agreement between European Member States over whether particular commercial releases of GM crops are environmentally "safe", despite a supposedly common approach to their risk assessment (Schomberg 1998;Wynne, Mayer 1999).Disputes routinely emerge over the appropriate scope of regulatory appraisal.Even where there is agreement over the possibility that effects will occur, notions of what constitute adverse effects remain strongly contested.
These sorts of problems with the current regulatory appraisal of GM crops are typical of those that beset the use of conventional risk assessment and cost benefit analysis in other areas.Taken together, these characteristics of the UK and European debate over the application of GM technologies in agriculture make it a challenging candidate for a case study concerning the potential for combining the systematic quantitative rigours of science with the breadth and flexibility of precaution.

The Multi-Criteria Mapping Pilot Study
The details of the multi-criteria mapping methodology and this particular application in the GM field in the UK are described at length elsewhere (Stirling, Mayer 1999, 2000).Only a summary will be attempted here.The study "A pilot multi-criteria mapping of a genetically modified crop in agricultural systems in the UK" was performed by the Science and Technology Policy Research (SPRU) unit of the University of Sussex and took place between April 1998 and September 1999.It was funded by the transnational food firm Unilever, but was entirely independent in its conception, design, implementation and reporting.The two authors come from an academic and NGO background and the project was overseen by a steering group comprising a wide range of environmental, consumer, farming and food industry representatives.The specific topic chosen for the study was the production of oilseed rape in the UK.Six "basic options" were identified in advance for the purposes of comparison (Table 1).

GM oilseed rape with post-release monitoring
Monitoring for effects (mainly environmental) conducted on an on-going basis after commercialisation.

GM oilseed rape with voluntary controls on areas of cultivation
Areas of growing of GM oilseed rape restricted on a voluntary basis to avoid unwanted effects such as geneflow and cross fertilisation of non-GM crops.
Up to six additional options to be specified by participant

Any option of participant's choice including combinations of the above if desired
In consultation with the steering group, twelve individual high profile protagonists in the GM food debate were selected as participants.These individuals came from a variety of backgrounds, including academics and government advisers, environmental, consumer and religious organisations and representatives from the farming, food and biotechnology industries.Of course, it is impossible to claim any "statistically representative" status for such a small sample.However, care was taken that, between them, the group of participants covered a full "envelope" of specialist and socio-political perspectives -ranging, for instance, from strongly opposed to strongly in favour of the use of GM crops.Given the character of the UK GM debate at the time of the study (1998)(1999), it was necessary to give undertakings of anonymity in order to secure participants' involvement.However, the viewpoints of participants are reproduced individually, being identified by a code letter and an affiliation with one of four general groupings of perspectives: "academic", "NGO", "industry" or "government" (Table 2).(iii) the scoring of the performance of each option under each criterion and (iv) the weighting of each criterion in terms of its relative importance.A straightforward "linear additive" multi-criteria procedure was employed using a simple spreadsheet model mounted on a portable computer in order to display to the participant the results of the scoring and weighting process as it developed.The process then iterated cyclically until the participant was satisfied that the results accurately reflected their own personal perspective and professional judgement on the issues at hand (Fig. 2).
One crucial feature of the scoring process was the deliberate eliciting of a range of performance scores for each option under each criterion.Rather than providing a single "best guess", this allowed consideration of a range of optimistic and pessimistic assumptions under each perspective.The discussion of the associated issues were then carefully documented and provided an interesting reflection of the importance of technical uncertainties in the final picture of performance derived under each perspective.
In the period following the interviews, participants were each contacted and asked to consider one final aspect of appraisal.This concerned the role of deliberate diversification among options as a means to implement two crucial features of the precautionary approach discussed earlier in this paper: the desire to hedge against intractable uncertainties and ignorance and to accommodate a variety of socio-political viewpoints.For this purpose, a simple index of diversity was borrowed from mathematical ecology and information theory in order to model different trade-offs between the value attached to the performance of individual options and the value attached to diversity in the mix of options.The details of this method are described in more detail elsewhere (Stirling, Mayer 1999, 2000).In short, participants can choose to assign a "zero", "low", "medium" or "high" weighting to diversity.Where this takes a zero value, then only the best-performing option is included in the favoured mix.Where higher weightings are placed on diversity, progressively greater contributions are included from lower-performing options -drawn on in proportion to their relative performance under the perspective in question.In this way, the study was able to elicit a crude notion of the importance of ignorance and pluralism under the different perspectives.
As analysis proceeded, participants were asked to review the results of a thorough sensitivity analysis conducted on their own set of weightings.Along with an invitation that participants retain the spreadsheet model and freely experiment with it at their leisure, this provided a further means to verify that the final results represented a robust reflection of the different viewpoints.Finally, all participants were invited to a concluding workshop at which the results were again confirmed and used as a basis for wider-ranging discussion.

Findings
The procedure summarised above generated a rich body of empirical material, reflecting a very wide range of issues bearing on the regulatory appraisal of GM technologies in agriculture.The key points will be summarised here.

Options
In addition to the six "basic options" a further 18 alternative agricultural strategies were identified and evaluated by participants.These included many different labelling and control regimes and the application of a series of different regulatory assessment procedures and criteria.They also involved a variety of different agricultural strategies, including a focus on hypothetical strategies using GM technologies under an organic farming regime.Interestingly these tended to perform relatively well under the perspectives of participants from both sides of the GM debate (Table 3).

Criteria
A total of 117 appraisal criteria were defined by different participants.Although some of these are apparently closely related, all embody important differences of emphasis or framing.
It is interesting that criteria typically reflect issues not only of consumer choice but also of citizenship and wider questions of participation and agency.Collectively they cover a very wide range of considerations, including envi-ronmental, agricultural, health, economic, social and ethical issues.Most of these issues are very remote from the factors included in orthodox risk assessment.Indeed, for no participant was it true that the full range of their concerns is fully addressed by existing regulatory appraisal procedures in the UK (Table 4)

Performance Scores
The pattern displayed by the performance scores assigned by participants under the various groups of criteria shows that the differences between perspectives are not simply due to variations in the weighting placed on different criteria.Participants adopt a variety of different "framing assumptions", resulting in significant differences in the scores assigned by different participants under the same criteria.This has implications for conventional multicriteria analysis, in which scoring is often conducted by a separate body of experts, with an assumption that different value judgements can be captured simply in the weightings.Despite these differences, however, the patterns evident in the scoring do allow certain general conclusions.For instance, under all perspectives the organic option tends to perform well under environmental criteria.Also, the evident divergence between the patterns of performance under health and environmental criteria challenge conventional assumptions that these aspects of performance will necessarily be well correlated.

Weightings
Notwithstanding the complications raised with respect to scoring above, the patterns evident in the weightings do illuminate some interesting, if rather unsurprising, features.Perspectives drawn from the biotechnology industry and food supply chain are conspicuous in their relative under-emphasis of the social and/or environmental and safety considerations which are prominent under all other perspectives.The perspectives adopted by government advisers have the distinctive characteristic of being at the same time relatively narrow in scope whilst emphasising environmental and safety considerations.The perspectives expressed by the non-industry participants share a markedly lower emphasis on economic or agricultural considerations.Seite 47

Rankings
Perhaps the most important output of the multicriteria mapping procedure are the patterns in the final rankings obtained under the various participant's perspectives.Indeed, it is this picture which constitutes the "map" referred to in the name of the technique, reflecting the way that final results vary with starting assumptions.Figure 3 displays the results obtained in the present exercise for the ten participants from whom it was possible to elicit all the necessary quantitative inputs (the remaining two -G and H -expressed difficulties either in the scoring or weighting process).Each chart represents the rankings elicited from a single participant.The sequence of basic options is the same in each case, running top to bottom from organic farming, through integrated pest management and conventional intensive agriculture to GM crops with (respectively) segregation and labelling, monitoring and voluntary controls.The extension of the horizontal bars to the right reflects the overall performance of the option under that perspective -the further to  the right, the better the performance.The scale is in arbitrary subjective linear units of performance.The length of the bars shows the difference between the performance under optimistic and pessimistic assumptions under each perspective, providing a useful indication of the relative importance of uncertainties.
A number of features can be distinguished in this picture.First, it is clear that there exist profound differences between the practical implications of the perspectives adopted by different participants.In an orthodox regulatory appraisal procedure (such as risk assessment) such variability would typically be concealed by the emphasis on a single aggregated position.In the present analysis, on the other hand, the idiosyncrasies of each individual position are clearly displayed.Second, the pervasive importance of uncertainty is shown by the overlap between the bars.However, it can be seen that the differences between ranking orders obtained under optimistic and pessimistic scores are generally rather small compared to the differences between perspectives.This suggests that it is not generally the technical dimensions of uncertainty which are the key issue, but rather more intangible qualitative aspects concerning the divergent interests, values and framing assumptions adopted by different participants.Third, despite the variability in the evaluations of different participants, there emerge a series of consistencies.With only a few exceptions, the organic option tends to perform relatively well.Conventional intensive agriculture tends to perform uniformly badly.The voluntary controls regime for GM crops is regarded rather poorly by industry, academic and NGO participants alike, being viewed favourably only by the government advisers.

Diversity
The final issue concerns the merits of diversification as a precautionary response to uncertainty, ignorance and divergent values.Diverse mixes of agricultural strategies were favoured by each of the seven participants who were able to respond to this part of the analysis.Each individual option relates to some arbitrary measure of importance, such as "share of output" or "share of land in production".For present purposes, the details do not affect the broad-brush conclusions.What is interesting is that six of these participants -drawn from all sides of the debate -favour the deliberate pursuit of some degree of diversity.The remaining participant -a senior UK government safety adviser -displays a uniquely high degree of confidence in the validity and robustness of their own perspective.Together with the idiosyncrasies evident in the rankings displayed in Figure 3, this is, in itself, an informative result.Taken as a whole, this part of the multi-criteria mapping analysis might be taken to raise questions over the extent to which R&D and regulatory policy making should be geared towards active encouragement of a variety of techniques rather than assuming or emphasising a single particular trajectory.Given the broad support expressed from all sides of the debate for the principle of diversity, questions might also be raised over how to treat options that display characteristics that militate against diversity.
In summary, then, this pilot multi-criteria mapping exercise permits a series of quite subtle and specific observations to be made concerning the character of the debate and the positions of the different protagonists.However, it also provides a basis for some general normative conclusions concerning the options under appraisal.These findings are all the more robust, for being set firmly in the context of systematic exploration and acknowledgement of the underlying variability and dissent.In particular, in the present case, the picture is quite unequivocal with respect to the relatively negative performance of conventional intensive farming and voluntary controls on the use of GM crops.If equal attention is paid to each perspective, then it is clear that the organic farming and integrated pest management options tend to perform significantly better overall.

Conclusions
This paper began with the identification of a set of eight criteria which might be held to combine key characteristics of both science and precaution.Although only a single pilot study, the present exercise does offer a basis for some conclusions over the practical efficacy of these criteria in the regulatory appraisal of technological risk.
The results of this exercise are more complete than orthodox risk assessment, in that they embody consideration of an unlimited array of issues.They also include consideration of a wide range of different strategic alternatives to the use of GM technologies.Equal attention is given to the justifications and benefits, as to the impacts and costs associated with the various options.The appraisal is inherently participatory, in that the results fully reflect the perspectives of a wide range of different socio-economic and cultural interests.The relative simplicity and auditability of the weighting and scoring approach retains a high degree of transparency.A central feature is the focus on the systematic mapping of the way in which specific results relate to specific assumptions in appraisal.This, together with the explicit attention given to more diverse mixes of options, embodies a greater degree of humility in the face of ignorance and pluralism than is normally displayed in risk assessment.
In this way, it may be concluded that there seems no reason in principle why conventional regulatory appraisal might not be adapted to better address the imperatives both of science and precaution.The present pilot exercise was not prohibitive in scale, being conducted by two researchers working part-time over a period of eighteen months.Though necessarily provisional, the consistencies displayed in the results show that it is possible -despite the breadth of the process -to draw some quite robust conclusions over the relative performance of options such as organic farming, conventional intensive agriculture and voluntary control regimes for the use of GM crops.Although the techniques of risk assessment continue to offer useful tools in addressing certain specific aspects, there seems to be no compelling theoretical, methodological or practical reasons for persisting in basing the entire business of regulatory appraisal on these inherently constrained and limited methods.Stirling, A., Mayer, S., 1999: Rethinking Risk: a pilot multi-criteria mapping of a genetically modified crop in agricultural systems in the UK.SPRU, University of Sussex Stirling, A., Mayer, S., 2000: A Novel Approach to the Appraisal of Technological Risk: a multi-criteria mapping pilot study of a genetically modified crop in the UK, Environment and Planning C, in press.Wynne, B., 1992: Uncertainty and Environmental Learning: reconceiving science and policy in the preventive paradigm.Global Environmental Change, pp. 111-127 Wynne, B., Mayer, S., 1999: The Release of Genetically Modified Organisms to the Environment.In "EU 98".European Environment Agency, Copenhagen

Figure
Figure 1: Risk, Uncertainty, Ambiguity and Ignorance

Figure
Figure 2: The Multi-Criteria Mapping Process

Figure
Figure 3: Final Rankings for the Basic Options under the Perspectives of Individual Participants : a.c.stirling@sussex.ac.ukInternet: http://www.sussex.ac.uk/spru/

Table 4 : Broad Groupings of Criteria Defined by Participants
Risk at a Turning Point?Journal of Risk Research, 1, 2, pp.97-110 Stirling, A., 1999: On Science and Precaution in the Management of Technological Risk.EUR 19056 EN, Seville: Institute for Prospective Technological Studies.This report draws on studies conducted by O. Renn and A. Klinke, A. Salo and A. Rip, as well as A. Stirling.