The Five Roles Framework, Revisited

This essay represents my most comprehensive engagement with disciplinary-specific AI to date. It offers the first published overview of existing approaches to disciplinary AI, in an effort to spark more collaboration and conversation, and it includes an extended re-examination of the specific roles in the Five Roles Framework. Most importantly, it answers the question at the heart of the framework: whether the roles function as a sequence or a routine.

This efforts will serve as the foundation for deeper integration and implementation of the Roles into specific lessons, units, and disciplines in subsequent articles and publications. This essay runs over 4,000 words, represents my best thinking to date, and, to my knowledge, is the first piece to work through the sequence-versus-routine question at any level of detail. For that reason, I’m releasing it only to paid subscribers.

This work takes real time, energy, and research to bring to you. Your support keeps it going.

In this article, I want to more deeply engage with my 5 Role Framework. Many readers, researchers, and instructors have found this framework helpful, but there still is a gap between vision and application that I want to recognize and clarify. Namely, does the 5 Role Framework function as a sequence or a repertoire of skills? This question matters because the desired relationship of the roles will have direct impacts on instructional design and practical implementation. If the roles function as a sequence, then classroom work should build around that sequence. If the roles function as a repertoire, then instructors can truly begin with any role in pursuit of discipline-specific academic outcomes.

First, let’s get a few basics out of the way. I have spent the past year developing the conception and practice of disciplinary-specific AI literacy (DSAIL). This approach hinges on the idea that general AI literacy only goes so far in preparing students “to think” in an AI-rich world. In reality, students need to see AI literacy play out in the specific contexts and domains that shape K-12 instruction, namely content areas or disciplines. In these domains, general AI literacy and disciplinary-specific epistemologies and practices circulate in a feedback loop. Students (and teachers) need to know enough about AI to engage with disciplinary-specific AI output and processes, even as disciplinary engagement activates and contextualizes that literacy content, simultaneously building deeper understanding of AI in pursuit of disciplinary goals and outputs. Here, the attentive reader can already see in embryonic form how I will answer my focus question.

But the above approach is still rather vague and hard to implement. So what if we can articulate the different epistemic commitments of specific disciplines? So what if each discipline uses specific and sometimes distinctive practices to develop and test knowledge? How does this inform the instructional introduction, adaptation, and evaluation of particular engagements with AI in disciplinary contexts? For a time, I was skeptical and critical of trans-disciplinary apparati that serve as an intermediary between general AI literacy and disciplinary-specific practice, but my work with districts and my DSAIL research cohort has underlined the necessity of such an intermediary. In essence, members of my cohort, while inspired by my general framework, through their own boots-on-the-ground experience and implementation, creatively developed intermediaries to enact a feedback loop between AI literacy and content-specific objectives and outcomes.