BAYESIAN NETWORKS

Bayesian Networks serve to systematically evaluate risk associated with a resource allocation task. A Bayesian Network is a graphical representation of “uncertain” variables (“nodes”) and their quantified dependencies. A Bayesian Network can serve, for example, for the assessment of risk connected with a product development. If an organisation wants to evaluate the probability of a successful market entry, the inter-dependent variables “Attitudes of Regulatory Agency”, “Safety of Product for User”, “Efficacy of Product” and “Technical Risk of Failure” may influence the central variable “Probability of Successful Market Entry”.

Core principal of a Bayesian Network is the possibility for the decision-makers to revise their initial (intuitively) risk assessments in the light of new information. Using a computer model, the decision-makers can systematically quantify all variables relevant for the overall risk assessment. For each node discreet “states” are defined: for example: “Favourable”, “Neutral”, “Unfavourable” for “Attitude of Regulatory Agency”. This quantification usually leads to a more precise and “objective” evaluation of the overall risk. The model is dynamic as parameters and therefore the overall risk assessment can be updated after obtaining new information.

Bayesian Networks can as well be used for a systematic evaluation of risk connected to projects of NGOs or governmental organisations.

Conditions to use Bayesian Networks to allocate resources:

• Success of decision is connected with a variety of dependent, uncertain variables
• Probabilities of the different states of the nodes can be estimated or calculated