Current understanding of the ecology of agroforestry practices, partic
ularly traditional practices in the tropics, is frequently qualitative
, sparse, and uncertain. Relevant information may be dispersed and com
plementary but not immediately compatible. Established approaches to d
ecision support, which tend to be deterministic, demand more precise b
ase information than is generally available for agroforestry practices
. As a result, they are of limited utility in helping development prof
essionals plan research and extension activities. In order to provide
decision support at an appropriate level, domain-specific software was
developed. This provides the user with an environment for creating kn
owledge bases by collating knowledge from a range of sources; facilita
tes the synthesis of that knowledge and its evaluation; and thereby fa
cilitates its use in planning agroforestry research and extension. The
software allows knowledge-base creation and exploration through text
or diagram interfaces and incorporates a suite of inference mechanisms
. A task language allows the combination of these mechanisms to be cus
tomized. This provides a powerful alternative to existing, less formal
approaches to evaluating the current state of knowledge on interdisci
plinary, problem-oriented topics as a basis for planning development a
ctivities.