This paper analyzes the use of collaborative problem solving techniques, which seek to provide a statistical framework for evaluating the effectiveness of the intervention of supplemental solvers. The presented approach is explained within a particular application domain, namely the routing of interconnection structures in ultralarge-scale integrated circuits. The paper starts by stating the inherent difficulties of the signal routing problem, and referencing the techniques that are currently used to tackle it. It continues by identifying some metrics that can be used to analyze the quality of the results produced by an automated router (the main problem solver), and identify the regions of the circuit that should be targeted for external intervention with custom interconnections (the supplemental solver). A statistical framework is then presented, which is intended to evaluate the effectiveness of the interventions of the supplemental solver, and to forecast the probability of success of further interventions.

The developed methods are not completely described in the paper, and readers are referred to three IBM research reports by the same authors for further details. The claims that the presented approach could be applied to collaborative problem solving in other application domains are also not directly addressed in the paper, and the authors admit in section 6 that their approach should be considered only a first attempt to formalize collaborative interactions among solvers. Taking that into account, the main value of the paper actually lies in the overview it provides on the research reports, about the design of on-chip interconnects. However, a better overview can be found in Wisniewski et al. [1].

As a final remark, the use of the term “collaborative” can be in some ways misleading, because the presented approach deals only with interactions among software tools, and not among users, as defined in the context of collaborative computing within the computer supported cooperative work (CSCW) area [2].