Argumentation with Evidence from Clinical Trials
Anthony Hunter (UCL) and Matthew Williams (UCH)
For more information, contact Anthony Hunter (a.hunter@cs.ucl.ac.uk).
Summary
Professionals need to make decisions, and increasingly they are turning to scientific evidence to help them. For example, clinicians use the results from clinical trials of drugs in helping them decide which drugs to use for a specific patient. The medical profession is at the vanguard of evidence-based decision-making, but it is being encouraged in other domains such as education, social work, policing and security, ecosystem management, and water management.
Since the scientific literature concerning any specific question may be very large and growing quickly, professionals turn to evidence-based guidelines that aggregate the evidence, and then through reasoned argument make recommendations for courses of action. Guidelines are written by teams of experts who collate and analyse the evidence in order to produce the recommendations.
As valuable as guidelines clearly are, there are shortcomings including: (1) Guidelines can quickly become out of date and so may be inconsistent with new evidence; (2) Multiple guidelines may apply and so lead to inconsistent recommendations; and (3) Guidelines do not take into account case-specific (contextual or personal) information.
In this project, we are developing technology for automatically aggregating evidence, and then producing recommendations based on the aggregated evidence. This will produce recommendations "on-the-fly", so that professionals can investigate the recommendations in terms of the evidence used to support them and the criteria used to aggregate the evidence. They can also obtain recommendations that are specific to a case, and that take account of the circumstances in which the decision is being made.
To illustrate our concerns, consider a clinician who has a patient with breast cancer and a long-standing chronic liver problem. Since these are regarded as disjoint medical problems, the available guidelines do not consider patients with both conditions. However, treatment of the breast cancer, e.g. chemotherapy or follow-on prophylactic treatment to reduce risk of reoccurrence may have serious negative ramifications on the chronic liver problem. The clinician then wants to re-evaluate the possible treatments to identify treatment options taking into account the trade-off of treating the cancer whilst not overly acerbating the liver problem. The system would draw on evidence such as clinical trials, observational studies, case reports, etc, that cover the negative ramifications of breast cancer treatments, and then it would re-evaluate the evidence concerning the efficacy of the breast cancer treatment in light of the potential negative ramifications. This would then lead to reasoned arguments, and possibly reasoned counterarguments, for what the best treatments would be for this patient. This process could harness the latest evidence and take into account any preferences that the clinician or patient may have.
Because scientific evidence is often complex, heterogeneous, incomplete and conflicting, argumentation is central to aggregating evidence and making recommendations So the project will develop a general theoretical framework for aggregation of heterogeneous knowledge based on a process of constructing and evaluating arguments and counterarguments so that knowledge is pooled to create strong and convincing arguments that recommend particular courses of action and so that significant conflicts in the available knowledge are highlighted and insignificant conflicts in the available knowledge are suppressed.
Links
Further information