Public awareness of genetic predisposition to diseases such as breast cancer threatens to put severe strain on genetics services. Computer-based decision support for general practitioners (GPs) has the potential to reduce unnecessary referrals, but issues of communicating about levels of risk and uncertainty must be addressed. An argumentation logic formalism can subsume both traditional probability theory and more qualitative, reason-based approaches to communicating uncertainty, and we propose that qualitative, argument-based presentation will make uncertainty information more accessible and comprehensible to both patient and GP. We describe software that uses an argumentation approach to assess genetic risk during a GP consultation and provide referral advice along with detailed qualitative explanations for its advice. The software was evaluated in real-life GP consultations in which actors played patients concerned about genetic risk, and in use by GPs evaluating simulated cases. Significant improvement in accuracy of assessment and appropriateness of referrals was found. GPs viewed the software and the qualitative reporting approach highly favourably.