“Pulling it All Together” via Psychometric AI

Rather long ago, Newell (1973) wrote a prophetic paper that could serve as a rallying cry for this 2004 symposium: “You Can’t Play 20 Questions with Nature and Win.” A number of those concerned with integrated cognition know of this paper, which helped catalyze both modern-day computational cognitive modeling through cognitive architectures (like ACT-R, Soar, Polyscheme, CLARION, etc.), and AI’s attempt to build a chess-playing machine better at the game than any human. However, not many know that in this paper Newell suggested a third avenue for achieving integration, one closely aligned with psychometrics. In the early days of AI, at least one thinker started down this road for a time (Evans 1968), but now the approach is long dead and all but forgotten. We recommend resurrecting this approach, in the form of what we call Psychometric AI, or just PAI (pronounced to rhyme with “π”). We briefly describe and defend PAI herein. We include some coverage of PERI, a robot in our lab who exemplifies PAI and integrated (artificial) cognition. We also explain how it is that we can strive for integration via the near-exclusive use of mechanized logic, under the umbrella of Newell’s third route.