Introduction

There are numerous works on food safety addressing risk governance in the contexts of specific crises such as BSE or emerging new problems such as genetically modified organisms (GMOs). On the other side, social studies of science developed focused analyses on uncertainty in scientific judgment. However, very few empirical studies look at the entire process, from risk assessment (RA) through risk management (RM) to public response from the perspective of uncertainty of expert knowledge and fewer still take an explicitly comparative approach. We conducted a European Union – United States comparison that investigated the ways scientific uncertainty is handled and presented by RAs, and RM are influenced by the amount and type of uncertainty found in the RAs that they respond to. At the RA phase, based on existing research and scientific principles experts must establishe what is certain about a particular food risk. Even when the same methods are used, standards of proof, and the quality or amount of research that settles disputes could vary depending on practicalities, regulations and epistemic cultures of the scientific communities involved.

Our project aimed at understanding the nature of scientific uncertainty in food safety science and its consequences for policy.

Approach

Our approach to scientific uncertainty is not normative, but empirical and comparative. We want to describe and understand how scientists express uncertainty in scientific reports assessing food risk. We look at English language risk assessment documents produced for the food safety agencies in the United States and the European Union between 2000 and 2010. We investigate two main, distinct areas of food hazards: contaminants and biohazards.

We coded the documents based on two ontologies of uncertainty and judgment (linguistic approach). The coded data are stored in a web database allowing analysis.  Then we tested the robustness of these ontologies using machine learning.

Main results

Concerning the nature of the expressed uncertainties, we highlighted three main points:

• National styles: risk assessments in the US contain more uncertainties and more precaution than the European documents
• Epistemic cultures: risk assessments of contaminants contain more uncertainties than the ones of biohazards, in both countries
• Knowledge accumulation: more knowledge does not reduce the uncertainty, but increases it. Our analysis reveals that the more time passed since the first cited RA on the subject and the more studies were referenced by the authors, the more uncertainty was expressed in their report. We also found that the more time the experts had to prepare the RA, the more uncertainty they conveyed. These seem to suggest that uncertainty is not diminishing with advances in knowledge but becomes more specific.

The relationship between quantity of uncertainty in risk assessment and the type of risk management decision is very weak and the recommendations written by risk assessors are not good predictors of the decision taken.