Composing with AI

Active Learning as a Pedagogy of Empiricism

Active learning serves as an adaptation of problem-posing pedagogy that composition process pedagogy fits well. If AI disrupts ‘writing about’ things, we must remember that we teach students ‘writing for’ purposes, audiences, and contexts, connecting the personal to the general and the past to the future, ideally. Merrill (2002) establishes active learning as a cycle of problem-posing, connecting the problem to life, modeling, application, and uptake:

1. Communicate in real-world problems: lessons and projects are based in visible issues and problems as opposed to abstract ideas or mechanics.
2. Activate previous experience: real-world problems for lessons should connect to students’ previous lived-experience or some existing shared knowledge to reconcile new material with students’ experience.
3. Demonstrate new material: teachers should model ways students can address the problem.
4. Require learners to apply new material and skills they have seen demonstrated: application enables assessment, coaching, and increasing challenge.
5. Integrate new knowledge into daily living: acquisition of new knowledge or skill is best indicated when it becomes associated with new lived-experiences (i.e. students can demonstrate, defend, or reflexively criticize it). (Merrill 2002 pp. 45–50)

In active learning, students are taught to identify problems, observe and collect related data, form and identify patterns in those observation, and build new experiences connected what they already do and know through testing the results in a safe environment, effectively connecting their past and present to consciously form a future. This process maps onto composition pedagogy beyond the parallels to Freirian problem-posing pedagogy as collaborative accumulation of critical experience (1970 p. 80).

Any composition assignment that requires students to address a mix of familiar and intentionally selected new experiences, such as readings and techniques, holds the potential to be an active project. Social epistemic pedagogy views composition as an empirical dialectical synthesis involving the writer, the community, and broader material conditions. It emphasizes the influence of the past on the present and future experiences of the writer. The aim is for students to "extraordinarily reexperience the ordinary" (Shor as cited in Berlin 1988 p. 681, Berlin 1982 p. 774). Post-process pedagogy encourages students to consider writing as always situated within a dynamic system, with implications for its circulation (McComiskey 2000). Transfer research in composition broadly advocates for students to think 'like writers' who can adapt their skills across different contexts, rather than being confined by the genres specific to their disciplines (Yancey, Robertson, and Taczak 2014 p. 9). It emphasizes engaging previous experiences in assignments that present expansive problems, rather than funneling students toward a single solution (Wardle 2012). These approaches, and others in composition pedagogy, align with active learning because they emphasize writing as a way to engage with an evolving world, rather than just packaging and transferring ideas from other fields or disciplines.

Active learning interprets education as knowledge-making: general “information” brought together with situated “portrayals” of limited knowledge to show problems, planning, and action in the context of a bigger picture that learners can navigate by learning/making knowledge (Merrill 2002 p. 48). In this sense, active learning is both empirical and rhetorical: noticing a problem and seeking to solve it requires gathering data about the world and the problem, identifying stakeholders, and arranging arguments combining those observations with other ecological imperatives to move toward actions that effect the situation (Bitzer 1968, Locke 1689, Cicero 2001, Edbauer 2005).

The connection between writing, empiricism, communication, and education sheds light on the challenge presented by AI writing tools and digital collaboration. If we view knowledge creation (problem-solving) as a natural outcome of communication, then enhancing knowledge production and problem-solving becomes as straightforward as improving communication technology. This perspective assumes that the nature of communication, including its rhetorical form and content, is inconsequential because it will all benefit from enhanced technology. This viewpoint is rooted in the Shannon-Weaver model of mathematical communication, which served as the foundation for Information Theory. Information Theory introduced the concept of the "bit," now widely recognized, as a means to translate qualitative information into technology-friendly binary mathematics, facilitating communication across distances and time (Gleick 2011 p. 229). Building on the idea that a coherent, informative sentence can be broken down into bits, the late-20th century Informational turn proposed that seemingly disorganized or unintelligible collections of observations, what we now refer to as ‘big data,' might inherently reveal valuable insights when treated like bits (Gleick 2011 pp. 230, 246). Information Science labels this process 'signal finding,' a common metaphor in contemporary empirical research, referencing the Shannon-Weaver model's capacity to encode signals so they can be 'detected' by intended recipients amid the 'noise' of other concurrent signals (Gleick 2011 p. 223). Data Science, a descendant of Information Theory, broadly involves using machines to extract patterns and signals that human labor cannot discern (Kelleher and Tierney 2018 pp. 1, 4). It places machines at the forefront of information and knowledge production, guided by human oversight.

GenAI also descends from Information Theory, using algorithms (complicated instructions) and machine learning (algorithms for pattern-extraction) to produce just-in-time information based on a user prompt (Gleick 2011 p. 57, Kelleher and Tierney 2018 p. 1). Information Theory, Data Science, and Machine Learning are guided by the DIKW Pyramid, a conceptual model illustrating knowledge ‘levels’ that humans and machines can both understand (Kitchin 2014 p. 10). “DIKW” stands for Data, Information, Knowledge, and Wisdom, each level of the pyramid (figure 1) building upon the preceding one through a distillation process (Kitchin 2014).

The DIKW pyramid symbolizes the transformation of noise (data) into a signal (information) and so on. This chapter will focus on these mechanics going forward. GenAI, akin to data science, uses DIKW logic by drawing on a training database (data) to identify patterns and generate outputs based on user input/prompts. By rhetorically exploring the mechanics of this pyramid, we can not only teach it to students as part of information literacy and argumentative writing but also gain insights into the ‘writing process’ of GenAIs and draw comparisons to students' own process labor.