The Contemporary Craft of Creating Characters Meets Today's Cognitive Architectures: A Case Study in Expressivity

Introduction

Ibsen’s characters are deep, memorable, and often dark; a contemporary Norwegian writer, Ullmann, is following suit.¹ With such fiction in mind, one might ask:

• Q1. Could an AI generate robust and engaging characters like these?

• Q2. Could revolutionary AI and sensory-immersion technology augment the ability of human writers to generate robust and engaging characters like these?

Elsewhere, answers, at least provisional ones, have been offered to this pair. In the case of (Q1), S. Bringsjord (2001) has defended a negative (a view not necessarily in concord with those of Licato); in the case of (Q2), Bringsjord & Bringsjord (2009) have defended an affirmative.² In the present chapter, the interest is in a very different question, one firmly rooted in the here and now, to wit:

• Q3. What does the contemporary craft of character design (by human authors), which is beyond the reach of foreseeable AI, and which isn’t powered by any stunning, speculative, AI-infused technology (immersive or otherwise), but is instead aided by tried-and-true “AI-less”³ software tools and immemorial techniques that are still routinely taught today, imply with respect to today’s computational cognitive architectures?

(Q1) is a question for AI researchers and engineers. (Q2) is a question for them as well, and for those working with them, from the field of human-computer interaction. (Q3) however, is not a question for engineers: it is a question for computational cognitive scientists, or more specifically, for computational cognitive modelers. (Q3) is relevant to those who proclaim, today, that they have on hand a “computational cognitive architecture” that captures, in one framework, most, if not all, of the nature and range of human cognition.

(Q3) can be concretized by pinning down the craft in question, and by doing the same for computational cognitive architectures and the modeling made possible by them. For the former, the authors accomplish this by focusing on the craft of character creation and design in the widely used, tried-and-true Movie Outline® system.⁴ For the latter, we focus on two cognitive architectures: ACT-R (Anderson, 1993, Anderson & Lebiere, 1998), probably the best-known cognitive architecture today, and one unmistakably aligned with the sanguine view that this architecture is well on its way to capturing the nature and breadth of human cognition (e.g. see the bold Anderson & Lebiere, 2003); and on CLARION (Sun, 2002), a cognitive architecture that uniquely founds its expressive power (among other things) on sub-symbolic processing, in keeping with what Sun (2001) has called “the duality of mind.”

The foregoing sets our general context. Since (Q3) is a large question, the authors narrow its scope so as to be able to productively consider it in the space of but the present chapter. The narrowing is accomplished by making three moves. First, the discussion is anchored to a simple, single story – Double-Minded Man – populated by only a pair of characters, and crafted in a particular “AI-less” software system taught and used by scriptwriters today. Second, the specific focus on the expressivity of a cognitive architecture, in light, specifically, of two expressivity challenges that arise from considering Double-Minded Man and its characters.⁵ Finally, cognitive architectures are partitioned, with respect to these challenges, into three categories. The first category is composed of those architectures that appear to have no chance of meeting the challenges; the second, of those that perhaps meet the challenges; and the third, of those that clearly meet the challenges. As shall be seen, the cognitive architecture ACT-R falls into the second category, and CLARION falls into the third. Encapsulating, then, here’s the core of our answer to (Q3):