Gambling on visions

As life becomes more complex, understanding its new risks requires greater ingenuity. Eva Pascoe explains. Suddenly, people have to deal with new, difficult problems such as DNA-based testing for breast cancer, the safety of genetically engineered food or risks of cancer from living close to petrol stations. Each problem is complex and full of uncertainty.

Risks to life and limb are inherent to new developments in many fields and systematic strategies for assessing and communicating risks are urgently required. Risk communication deals with advancing knowledge about risk and is central to the management of the impact of new technologies, such as biotechnology or food irradiation. A report on Risk analysis, perception and management by the Royal Society in 1992, and the Health and Safety Executive have recognised the importance of risk communication.

People's ability to respond to environmental and food hazards is determined, in part, by their understanding of the processes that govern its creation and control. Lacking scientific evidence, people often have to make educated guesses, based on whatever they know about a hazard; this may lead to misperception and confusion.

Research has shown that accurate knowledge about new food technology, such as food irradiation, translates into greater acceptance. The arguments and evidence should be presented transparently, with the uncertainty and limits in knowledge explicitly conveyed. Such arguments would be composed of many parts, which may lead to difficulties in understanding. A good format for communication of complex, risky issues should fulfil not only the basic criteria of correctness but also comprehensibility of the information. Most formats for communicating new risks use conventional, linear text. Yet, research shows that people find it difficult to understand linear arguments presented in lines of solid text.

The main reported problems are the difficulties people have in identifying the structure of arguments in terms of differentiating between main problems and supporting evidence. People also tend to over-emphasise novel arguments.

These basic problems indicate that text-based linear arguments may not be the best way of presenting information on "risky" issues. Text is not easy to understand and may need to be supported by other formats to produce an easy, user-friendly interface. Additionally, the apparent difficulties in complex argument processing make risk information particularly susceptible to modification by people with different values, with some groups presenting an emphasis on arguments for, while others emphasise the argument against. For example, a consumer group versus industry spokesman typically involves both groups presenting carefully selected evidence to support two opposite sides of the same argument. As concluded by Wald and Doll, "one rule is absolute; that all the available evidence must be taken into account".

Also, it has been indicated that people have difficulties due to the technical vocabulary and phraseology in risk information. Using hard words which could lead to misconceptions such as reading specificity versus sensitivity in testing, using hard to envision processes (tumour development, the process of photosynthesis), or using ideas that are hard to understand because they are hard to believe (that wholesome foods may contain natural carcinogens at higher levels than humanly produced carcinogenic pesticide residues) has led to misunderstandings and misinterpretation of the risk information. At present most of the work on risk communication offers practitioners helpful guidelines such as "speak clearly", "avoid jargon" and "speak with compassion". While important, these guidelines have limitations as they help in determining why a communication failed.

Poor risk communication can lead to wrong decisions by omitting key information or failing to contradict misconceptions, create confusion by making inappropriate assumptions. This often leads to conflict.

To develop a framework to deal with the difficulties that occur when complex arguments are presented in linear text, the Standardised Argument Report, StAR, project was set up in October 1993, with the Imperial Cancer Research Fund and City University as partners. This is an opportunity for cognitive psychologists to put our ideas into practice and join forces with expert system developers to provide easy and transparent communication of risk via computers.

The aim is the development of a computer-based risk assessment tool to help toxicologists evaluate the carcinogenicity of new compounds. We are using qualitative reasoning AI techniques. The project will provide a product, which will serve experts, novice toxicologists or chemists with a more general background. The way the information is shown on the screen should enhance understanding and make it more comprehensible and unambiguous.

A typical argument in linear text-only form for the lay-public, on the safety of genetically engineered food such as transgenic fish, focusses on the safety of the gene insertion process (Berkowitz, 1994): "The remaining question concerns the possibility of unexpected toxic effects arising as a result of the insertion of a transgene into the host genome. In fish, one such possibility is an insertion causing the expression of a quiescent toxin gene, in a normally safe species of fish. Some fish species contain toxins, and animals near the superclass Pisces in the evolutionary hierarchy contain toxins. Therefore, one might suspect that species of fish normally known to be safe may carry unexpressed genes coding for toxic polypeptides or for enzymes that synthesise chemical toxins, and that these genes may be activated by genetic insertions. However, we believe that the probability of activating a toxin gene is very small, because normally safe fish are very unlikely to have toxin genes that can be activated".

Our approach is to use multimedia technology to eliminate the emphasis on text. We intend to produce a communication format that will convert the linear text argument into a more transparent, graphical representation indicating main arguments and supporting reasons.

StAR technology will provide tools to visualise the argument generated by an expert system during the risk characterisation phase. Multimedia information will accompany definitions, illustrations and explanations of difficult words and phrases like "genetic insertion" and "unexpressed genes". Further explanation and details of references to complex words and processes could be obtained by clicking on appropriate sections of the argument visualisation.

Hypertext links to textual information on a CD-Rom of reference information will provide the facility to go deeper into a given problem or concept. Text will be used as one of the search vehicles, but in combination with a multimedia representation of processes and concepts.

The project is sponsored by DTI/ESRC Intelligent System Integration Programme, and involves a number of organisations and people. John Fox from the Imperial Cancer Research Fund leads the team of artificial intelligence experts including Paul Krause, ICRF, and cognitive psychologists Peter Ayton and myself at the City University. The project is managed by Phillip Judson and software is supplied by Logic Programming Associates.

Our visualisation of difficult arguments will provide a tool for toxicologists and also help in other areas where uncertainty and complexity has hindered public participation in decision-making, such as the Maastricht Treaty or Criminal Justice Bill.

Clear and easy presentation of arguments would help experts and the general public develop an understanding of the new problems and risks involved. Complex problems, new technologies and uncertainty caused by them are here to stay. We need more projects like StAR to use multimedia technology to provide transparent ways of communicating complex and uncertain issues.

Eva Pascoe is a cognitive psychologist at City University, London.