XOMO: Understanding Development Options for Autonomy

We seek an AI agent that is fast enough to keep up with debates between humans and and can offer suggestions regarding what is the next most important issue to explore. To assess the merits of treatment learning for such an agent, this kind of learning was applied to three models of a mythical (but plausible) development project building software for autonomous systems. The models were the COCOMO effort estimation model; the COQUALMO defect introduction and removal model; and Madachy’s Heuristic Schedule Risk model. The experiment was successful; i.e. the learner found ways to improve reduce the residual defects per thousand lines of code by 85% while halving the risk that the schedule over run. Also, the development effort was nearly halved. Hence, we conclude that treatment learners can understand how to improve the process of building autonomous software.