Press "Enter" to skip to content

Tag: online education

The Ghost in the Feedback Loop: AI, Academic Praxis, and the Decomposition of Disciplinary Boundaries

The following are the slides and synopsis of my paper, The Ghost in the Feedback Loop: AI, Academic Praxis, and the Decomposition of Disciplinary Boundaries, presented at the International Society for the Scholarship of Teaching and Learning Annual Conference (ISSOTL 2025), in the University of Canterbury, Christchurch, New Zealand.

Eldritch Technics | Download PDF

As AI tools transform content creation, academic practices, and disciplinary boundaries are under pressure. Drawing on Actor-Network Theory (ANT), this paper explores AI tools as nonhuman actants shaping authorship, assessment, and pedagogical authority (Fenwick & Edwards, 2010, 2012). ANT challenges humanist binaries such as human/machine by inviting us to view education as an assemblage of human and nonhuman actors co-constructing the learning environment (Landri, 2023).

Within this framework, AI systems used in formative assessment, ranging from feedback automation to individual AI tutoring, reshape pedagogic feedback loops, influence student agency, and reconfigure the distribution of cognitive labor in classrooms (Hopfenbeck et al., 2024; Zhai & Nehm, 2023). As students increasingly co-produce knowledge with AI (Wang et al., 2024), this paper argues that the pedagogical focus must shift from control and containment to composition and negotiation. Using case studies from large international cohorts, the paper examines how AI alters feedback loops, shifts student agency, and challenges discipline-specific praxis. What new academic identity and ethics forms must emerge in this hybrid landscape?

Recent studies suggest that generative AI can reduce perceived cognitive effort while paradoxically elevating the problem-solving confidence of knowledge workers (Lee et al., 2025). When strategically embedded in formative assessment practices, AI can scaffold students’ movement up Bloom’s taxonomy from comprehension to application, analysis, and synthesis, especially among international and multilingual cohorts (Walter, 2024; Klimova & Chen, 2024).

In this context, this paper argues for a radical reframing of educational assessment design. Instead of resisting machinic participation, educators must critically reassemble pedagogical networks that include AI as epistemic collaborators (Liu & Bridgeman, 2023). By unpacking the socio-material dynamics of AI-infused learning environments, ANT offers a pathway for understanding and designing inclusive, dynamic, and ethically aware pedagogical futures. This includes rethinking agency as distributed across human and nonhuman nodes, assessment as an ongoing negotiation, and learning environments as fluid, adaptive ecologies shaped by constant assemblage and reassemblage rather than fixed instructional designs or isolated learner outcomes.

References
Fenwick, T., & Edwards, R. (2010). Actor-Network Theory in Education. Routledge. https://doi.org/10.4324/9780203849088

Fenwick, T., & Edwards, R. (Eds.). (2012). Researching Education Through Actor-Network Theory. Wiley-Blackwell. https://doi.org/10.1002/9781118275825

Hopfenbeck, T. N., Zhang, Z., & Authors (2024). Challenges and opportunities for classroom-based formative assessment and AI: A perspective article. International Journal of Educational Technology, 15(2), 1–28.

Klimova, B., & Chen, J. H. (2024). The impact of AI on enhancing students’ intercultural communication, competence at the university level: A review study. Language Teaching Research Quarterly, 43, 102-120. https://doi.org/10.32038/ltrq.2024.43.06

Landri, P. (2023). Ecological materialism: redescribing educational leadership through Actor-Network Theory. Journal of Educational Administration and History, 56, 84 – 101. https://doi.org/10.1080/00220620.2023.2258343.

Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Proceedings of the ACM CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3544548.3581234

Liu, D. & Bridgeman, A. (2023, July 12). What to do about assessments if we can’t out-design or out-run AI? University of Sydney. https://educational-innovation.sydney.edu.au/teaching@sydney/what-to-do-about-assessments-if-we-cant-out-design-or-out-run-ai/

Walter, Y. (2024). Embracing the future of artificial intelligence in the classroom: The relevance of AI literacy, prompt engineering, and critical thinking in modern education. International Journal of Educational Technology in Higher Education, 21, Article 15. https://doi.org/10.1186/s41239-024-00448-3

Wang, S., Wang, F., Zhu, Z., Wang, J., Tran, T., & Du, Z. (2024). Artificial intelligence in education: A systematic literature review. Expert Syst. Appl., 252, 124167. https://doi.org/10.1016/j.eswa.2024.124167

Zhai, X., & Nehm, R. H. (2023). AI and formative assessment: The train has left the station. Journal of Research in Science Teaching, 60(6), 1390–1398. https://doi.org/10.1002/tea.21885

On digital artefacts

I recently recorded 3 short YouTube lectures on developing digital artefacts. Previously, I’ve outlined how I use the concept of digital artefacts in my piece on teaching digital media in a systemic way. In brief, the digital artefact assessment framework I developed gives students the opportunity to work on projects with real-world implications and relevance, that is, projects with nonlinear outcomes aimed at real stakeholders, users, and audiences. The only criteria for a digital artefact are that 1] artefacts should be developed in public on the open internet, therefore leveraging non-linearity, collective intelligence and fast feedback loops, and 2] artefacts should have a clearly defined social utility for stakeholders and audiences outside the subject and program. Below is my lecture outlining the key aspects of the digital artefact development process.

Fail Early, Fail Often [#fefo] is a developmental strategy originating in the open source community, and first formalized by Eric Raymond in The Cathedral and the Bazaar.  In the context of teaching and learning, FEFO asks creators to push towards the limits of their idea, experiment at those limits and inevitably fail, and then to immediately iterate through this very process again, and again. At the individual level the result of FEFO in practice is rapid error discovery and elimination, while at the systemic level it leads to a culture of rapid prototyping, experimentation, and ideation. Below is my lecture outlining the use of the #fefo philosophy in developing digital artefacts.

Fast, Inexpensive, Simple, Tiny [#fist] is a developmental strategy developed by Lt. Col. Dan Ward, Chief of Acquisition Innovation at USAF. It provides a rule-of-thumb framework for evaluating the potential and scope of projects, allowing creators to chart ideation trajectories within parameters geared for simplicity. In my subjects FIST projects have to be: 1] time-bound [fast], even if part of an ongoing process; 2] reusing existing easily accessible techniques [inexpensive], as opposed to relying on complex new developments; 3] constantly aiming away from fragility [simple], and towards structural simplicity; 4] small-scale with the potential to grow [tiny], as opposed to large-scale with the potential to crumble. Below is my lecture outlining the use of the #fefo methodology in developing digital artefacts.

Teaching digital media in a systemic way, while accounting for non-linearity

Recently I have been trying to formulate my digital media teaching and learning philosophy as a systemic framework. This is a posteriori work because philosophies can be non-systemic, but systems are always based on a philosophy. I also don’t think a teaching/learning system can ever be complete, because entropy and change are the only givens [even in academy]. It has to be understood as dynamic, and therefore more along the lines of rules-of-thumb as opposed to prescriptive dogma.

None of the specific elements of the framework I use are critical to its success, and the only axiom is that the elements have to form a coherent system. By coherence, I understand a dynamic setting where 1] the elements of the system are integrated both horizontally and vertically [more on that below], and 2] the system is bigger than the sum of its parts. The second point needs further elaboration, as I have often found even highly educated people really struggle with non-linear systems. Briefly, linear progression is utterly predictable [x + 1 + 1…= x + n] and comfortable to build models in – i.e. if you increase x by 1, the new state of the system will be x +1. Nonlinear progression by contrast is utterly unpredictable and exhibits rapid deviations from whatever the fashionable mean is at the moment – i.e. x+1= y. Needless to say, one cannot model nonlinear systems over long periods of time, as the systems will inevitably deviate from the limited variables given in the model.

Axiom: all complex systems are nonlinear when exposed to time [even in academy].

The age of the moderns has configured us to think exceedingly in linear terms, while reality is and has always been regretfully non-linear [Nassim Taleb built a career pointing this out for fun and profit]. Unfortunately this mass delusion extends to education, where linear thinking rules across all disciplines. Every time you hear the “take these five exams and you will receive a certificate that you know stuff” mantra you are encountering a manifestation of magical linear thinking. Fortunately, learning does not follow a linear progression, and is in fact one of the most non-linear processes we are ever likely to encounter as a species.

Most importantly, learning has to be understood as paradigmatically opposed to knowing facts, because the former is non-linear and relies on dynamic encounters with reality, while the latter is linear and relies on static encounters with models of reality.

With that out of the way, let’s get to the framework I have developed so far. There are two fundamental philosophical pillars framing the assessment structure in the digital media and communication [DIGC] subjects I have been teaching at the University of Wollongong [UOW], both informed by constructivist pedagogic approaches to knowledge creation [the subjects I coordinate are BCM112, DIGC202, and DIGC302].

1] The first of those pillars is the notion of content creation for a publicly available portfolio, expressed through the content formats students are asked to produce in the DIGC major.

Rule of thumb: all content creation without exception has to be non-prescriptive, where students are given starting points and asked to develop learning trajectories on their own – i.e. ‘write a 500 word blog post on surveillance using the following problems as starting points, and make a meme illustrating your argument’.

Rule of thumb: all content has to be publicly available, in order to expose students to nonlinear feedback loops – i.e. ‘my video has 20 000 views in three days – why is this happening?’ [first year student, true story].

Rule of thumb: all content has to be produced in aggregate in order to leverage nonlinear time effects on learning – i.e. ‘I suddenly discovered I taught myself Adobe Premiere while editing my videos for this subject’ [second year student, true story].

The formats students produce include, but are not limited to, short WordPress essays and comments, annotated Twitter links, YouTube videos, SoundCloud podcasts, single image semantically-rich memetic messages on Imgur, dynamic semantically-rich memetic messages on Giphy, and large-scale free-form media-rich digital artefacts [more on those below].

Rule of thumb: design for simultaneous, dynamic content production of varying intensity, in order to multiply interface points with topic problematic – i.e. ‘this week you should write a blog post on distributed network topologies, make a video illustrating the argument, tweet three examples of distributed networks in the real world, and comment on three other student posts’.

 2] The second pillar is expressed through the notion of horizontal and vertical integration of knowledge creation practices. This stands for a model of media production where the same assessments and platforms are used extensively across different subject areas at the same level and program of study [horizontal integration], as well as across levels and programs [vertical integration].

Rule of thumb: the higher the horizontal/vertical integration, the more content serendipity students are likely to encounter, and the more pronounced the effects of non-linearity on learning.

Crucially, and this point has to be strongly emphasized, the integration of assessments and content platforms both horizontally and vertically allows students to leverage content aggregates and scale up in terms of their output [non-linearity, hello again]. In practice, this means that a student taking BCM112 [a core subject in the DIGC major] will use the same media platforms also in BCM110 [a core subject for all communication and media studies students], but also in JOUR102 [a core subject in the journalism degree] and MEDA101 [a core subject in media arts]. This horizontal integration across 100 level subjects allows students to rapidly build up sophisticated content portfolios and leverage content serendipity.

Rule of thumb: always try to design for content serendipity, where content of topical variety coexists on the same platform – i.e. a multitude of subjects with blogging assessments allowing the student to use the same WordPress blog. When serendipity is actively encouraged it transforms content platforms into so many idea colliders with potentially nonlinear learning results.

Adding the vertical integration allows students to reuse the same platforms in their 200 and 300 level subjects across the same major, and/or other majors and programs. Naturally, this results in highly scalable content outputs, the aggregation of extensively documented portfolios of media production, and most importantly, the rapid nonlinear accumulation of knowledge production techniques and practices.

On digital artefacts

A significant challenge across academy as a whole, and media studies as a discipline, is giving students the opportunity to work on projects with real-world implications and relevance, that is, projects with nonlinear outcomes aimed at real stakeholders, users, and audiences. The digital artefact [DA] assessment framework I developed along the lines of the model discussed above is a direct response to this challenge. The only limiting requirements for a DA are that 1] artefacts should be developed in public on the open internet, therefore leveraging non-linearity, collective intelligence and fast feedback loops, and 2] artefacts should have a clearly defined social utility for stakeholders and audiences outside the subject and program.

Rule of thumb: media project assessments should always be non-prescriptive in order to leverage non-linearity – i.e. ‘I thought I am fooling around with a drone, and now I have a start-up and have to learn how to talk to investors’ [second year student, true story].

Implementing the above rule of thumb means that you absolutely cannot structure and/or limit: 1] group numbers – in my subjects students can work with whoever they want, in whatever numbers and configurations, with people in and/or out of the subject, degree, university; 2] the project topic – my students are expected to define the DA topic on their own, the only limitations provided by the criteria for public availability, social utility, and the broad confines of the subject area – i.e. digital media; 3] the project duration – I expect my students to approach the DA as a project that can be completed within the subject, but that can also be extended throughout the duration of the degree and beyond.

Digital artefact development rule of thumb 1: Fail Early, Fail Often [FEFO]

#fefo is a developmental strategy originating in the open source community, and first formalized by Eric Raymond in The Cathedral and the Bazaar. FEFO looks simple, but is the embodiment of a fundamental insight about complex systems. If a complex system has to last in time while interfacing with nonlinear environments, its best bet is to distribute and normalize risk taking [a better word for decision making] across its network, while also accounting for the systemic effects of failure within the system [see Nassim Taleb’s Antifragile for an elaboration]. In the context of teaching and learning, FEFO asks creators to push towards the limits of their idea, experiment at those limits and inevitably fail, and then to immediately iterate through this very process again, and again. At the individual level the result of FEFO in practice is rapid error discovery and elimination, while at the systemic level it leads to a culture of rapid prototyping, experimentation, and ideation.

Digital artefact development rule of thumb 2: Fast, Inexpensive, Simple, Tiny [FIST]

#fist is a developmental strategy developed by Lt. Col. Dan Ward, Chief of Acquisition Innovation at USAF. It provides a rule-of-thumb framework for evaluating the potential and scope of projects, allowing creators to chart ideation trajectories within parameters geared for simplicity. In my subjects FIST projects have to be: 1] time-bound [fast], even if part of an ongoing process; 2] reusing existing easily accessible techniques [inexpensive], as opposed to relying on complex new developments; 3] constantly aiming away from fragility [simple], and towards structural simplicity; 4] small-scale with the potential to grow [tiny], as opposed to large-scale with the potential to crumble.

In the context of my teaching, starting with their first foray into the DIGC major in BCM112 students are asked to ideate, rapidly prototype, develop, produce, and iterate a DA along the criteria outlined above. Crucially, students are allowed and encouraged to have complete conceptual freedom in developing their DAs. Students can work alone or in a group, which can include students from different classes or outside stakeholders. Students can also leverage multiple subjects across levels of study to work on the same digital artefact [therefore scaling up horizontally and/or vertically]. For example, they can work on the same project while enrolled in DIGC202 and DIGC302, or while enrolled in DIGC202 and DIGC335. Most importantly, students are encouraged to continue working on their projects even after a subject has been completed, which potentially leads to projects lasting for the entirety of their degree, spanning 3 years and a multitude of subjects.

In an effort to further ground the digital artefact framework in real-world practices in digital media and communication, DA creators from BCM112, DIGC202, and DIGC302 have been encouraged to collaborate with and initiate various UOW media campaigns aimed at students and outside stakeholders. Such successful campaigns as Faces of UOW, UOW Student Life, and UOW Goes Global all started as digital artefacts in DIGC202 and DIGC302. In this way, student-created digital media content is leveraged by the University and by the students for their digital artefacts and media portfolios. To date, DIGC students have developed digital artefacts for UOW Marketing, URAC, UOW College, Wollongong City Council, and a range of businesses. A number of DAs have also evolved into viable businesses.

In line with the opening paragraph I will stop here, even though [precisely because] this is an incomplete snapshot of the framework I am working on.

My YouTube lectures

This semester I’ve started uploading my lectures for DIGC202 Global Networks to YouTube, while abandoning the face-to-face lecture format in that subject. The obvious benefit of this shift is to allow students to engage with the lectures on their own terms – the lectures are broken into segments which can be accessed discretely or in a sequence, on any device, at any time. The legacy alternative would have been either attending a physical lecture or listening to the university-provided recording, which is an hour-long file hidden within the cavern of the university intranet, accessible only from a computer [must keep that knowledge away from prying eyes!], and, as a rule of thumb, of terrible quality. Anecdotal evidence from students already validates my decision to shift, as this gives them the ability to structure their learning activities in a format productive for them.

The meta-benefit is that the lectures – and therefore my labour – now exist within a generative value ecology on the open net, accessible to [gasp] people outside the university. On a more strategic level, I can now annotate the lectures as I go along, adding links to additional content which will only enrich the experience. In that sense the lectures stop being an end-product, an artefact of dead labour [dead as in dead-end], and become an open process.

The only downside I have had to deal with so far is that lecture preparation, delivery, and post-production takes me on average three times as long as the legacy model. I am still experimenting with the process and learning on the go – fail early, fail often.

I am uploading all lectures to a DIGC202 playlist, which can be accessed below: