possible-thesis-chapters

the thesis feel so abstractly far in the future at the moment, but I tried the exercise of sketching out possible chapters and its helpful in making clear to myself the important corners of the research investigation

defining a playful methodology - drawing on speculative design, design fiction, and ~~play theory~~ live coding and performance (as playful)
process and performance - live coding as literal performance, but also the performative nature of many institutional practices
process - digital gardens, tools for thought, personal knowledge management and ‘second brains’
intelligence and knowledge - the problematic history of how intelligence has been defined and measured, and how that has shaped the development of AI over time
- power and ethics - both in commercial AI production and within academic institutions
possibility and (im)possibility spaces - exploring the concept itself, but also the constrained and difficult space of higher education as an institution.
- though experiments (speculative design) and the socio economic/ political context that shapes the work
autoethnographic element

1: introduction

Frames the research questions: what can AI add to learning and knowledge-making without eroding the generative space creative work depends on, and what assumptions about intelligence and agency are embedded in commercial AI systems entering education.
Introduces ‘After Intelligence’ as a layered framing: The temporal gesture past current hype, the critical interrogation of intelligence as a concept. Names the inversion from the original proposal (designing for others) to the current work (the researcher’s own practice as site and subject).
Introduces PhD-Live and the wider infrastructure briefly. Sets out the thesis structure and explains the distributed autoethnographic and pedagogical approach, so the reader knows what to expect.

2: methodology, a playful methodology

Conceptual argument:

Positions the research within practice-based and autoethnographic traditions, then builds a methodological framework through speculative design (thought experiments, Dunne and Raby), design fiction, establishing the idea of possibility spaces as speculative design practice that is being employed as foundational approach, which led to exploring live coding as a playful and performative practice that embodies liveness. Addresses liveness as a methodological commitment, not an aesthetic one.
Argues that ‘building with and through’ LLM systems is the primary mode of investigation, where each artefact is both technical object and theoretical proposition.
Names and justifies the structural decision to distribute pedagogy and autoethnographic reflection across all chapters rather than isolating them.

Pedagogical grounding: Positions the methodology in relation to critical and emancipatory pedagogy (Freire, hooks). The choice of a playful methodology is not neutral: it is a response to how learning gets flattened in institutional and commercial AI contexts. (play as way to learn sources Montessori approach)

Autoethnographic account: Reflects on occupying the triple role of learner, researcher, and teacher simultaneously, and what that does to methodological choices.

3: intelligence, power and ethics

Conceptual argument: Traces how intelligence has been defined, measured, and operationalised, and how that history has shaped the development of AI (Crawford, Agüera y Arcas, Hayles, critical AI studies, histories of psychometrics and classification). The chatbot lineage matters here: Weizenbaum built ELIZA as a critique not a product, and that parallel to building bots that interrogate rather than optimise is worth making explicit. Examines how these logics enter educational contexts through commercial AI systems, and what assumptions about agency, autonomy, and reasoning they carry. Introduces the bots here as critical responses: artefacts built to interrogate these assumptions rather than reproduce them. Addresses the political economy of commercial AI production and the power dynamics that determine how AI gets adopted within institutions and who gets to question it.

Pedagogical grounding: Definitions of intelligence determine what gets optimised in educational technology. Manning on learning otherwise and refusal of institutional categories: learning exceeds what intelligence frameworks measure. This chapter establishes the stakes before the thesis turns to knowledge.

Autoethnographic account: Through the experience of building tools that deliberately resist the resolve-and-optimise logic, reflects on what it means to construct alternative models of intelligence in practice. The echo chamber problem in the supervisor bot surfaces here, not as a technical finding (that is chapter 4) but as a consequence of how LLM “intelligence” actually works. The sycophancy literature (Cheng et al, Malmqvist) connects: these are structural tendencies in LLMs, not incidental failures. Through navigating institutional AI adoption as both staff and student, reflects on precarity and structural position as conditions that shape what questions feel askable.

4: process, digital gardens, tools for thought, and knowledge infrastructure

Conceptual argument: Reviews digital gardens, personal knowledge management, tools for thought (Matuschik, Nielsen, Appleton), and the “second brain” tradition. Positions PhD-Live within this lineage but argues it does something these frameworks don’t: it treats process visibility as a critical response to AI’s tendency to collapse process into product. Introduces the local research infrastructure (supervisor bot, study companion, shared context layer) as constructed tools for thinking, not autonomous agents. Documents what they are and how they work.

Pedagogical grounding: Tools for thought are fundamentally claims about how learning happens. This section examines whose model of learning these tools encode, and what happens when the learner is also the builder.

Autoethnographic account: Through the experience of building and using PhD-Live and the local infrastructure, deeper questions about learning surface. The echo chamber problem in the supervisor bot as a finding about how correlated context collapses the space for thinking. The capability gap between local models and Claude as a finding about the sovereignty/quality tradeoff. What it means to learn in public when the infrastructure is yours.

5: knowledge, power and ethics

Conceptual argument: Shifts from how intelligence is defined to what counts as knowledge and who gets to produce it. Draws on tacit knowledge traditions (Schon, Polanyi, Dreyfus) to argue that the forms of knowing central to creative practice resist the explicit, extractable model that LLMs embody. Examines how institutional and commercial knowledge systems encode judgements about what is legitimate, rigorous, or valuable. Connects to the politics of knowledge within higher education: whose knowledge counts, what gets credentialled, how AI systems interact with those existing hierarchies.

Pedagogical grounding: This is where Freire and hooks do their deepest work. Knowledge-making as a political act, not a neutral process. The question of how AI participates in creative thinking without speaking for the researcher is fundamentally a question about knowledge and power.

Autoethnographic account: Through the experience of making knowledge in public via PhD-Live while navigating institutional structures that assess and credential that knowledge, reflects on the tension between open, unfinished thinking and the institutional demand for legibility. The capability gap between local models and frontier systems surfaces here as a finding about knowledge-sovereignty: who controls the infrastructure of your own thinking, and what you give up to maintain that control.

6: possibility or (im)possibility spaces

Conceptual argument: What did the possibility spaces offer? What happened, a moment for some reflection on the speculation? But also impossibility: the constrained, material, political reality of higher education as an institution. The tension between what can be imagined and what can be built given actual resources, time, precarity, and institutional pressure. The decision to build locally rather than rely on commercial APIs as both a political and practical choice shaped by these constraints. This is a reflective, evaluative chapter: the speculative framing meets what actually happened.

Pedagogical grounding: What kinds of learning become possible or impossible depending on infrastructure, institutional context, and who controls the tools? The speculative move is not abstract: it asks what learning could look like if the conditions were otherwise.

Autoethnographic account: Through the experience of building within real constraints (time, capacity, institutional position, the limits of local hardware), reflects on what the impossibility spaces reveal about the conditions of knowledge-making that the possibility spaces alone cannot. The gap between what was designed and what was actually buildable becomes itself a finding about the political economy of creative research.

7: conclusion

Returns to the research questions. Synthesises the contributions across chapters. States the originality claim plainly: that the work treats building with and through AI as a method for investigating the conditions under which knowledge-making happens within a creative institution. Reflects on what “After Intelligence” means at the end of the project, having lived through it. Discusses limitations and future directions, including ICLC and the ongoing life of PhD-Live beyond the thesis.

Revision history 5 revisions

The revisions below show how this note has changed over time.

Frames the research questions: what AI can add to learning and knowledge-making without eroding the generative space creative work depends on, and what assumptions about intelligence and agency are embedded in commercial AI systems entering education. Introduces "After Intelligence" as a layered framing: the pun on AI, the temporal gesture past current hype, the critical interrogation of intelligence as a concept. Names the inversion from the original proposal (designing for others) to the current work (the researcher's own practice as site and subject). Introduces PhD-Live and the wider infrastructure briefly. Sets out the thesis structure and explains the distributed autoethnographic and pedagogical approach, so the reader knows what to expect. - Frames the research questions: what can AI add to learning and knowledge-making without eroding the generative space creative work depends on, and what assumptions about intelligence and agency are embedded in commercial AI systems entering education. - Introduces 'After Intelligence' as a layered framing: The temporal gesture past current hype, the critical interrogation of intelligence as a concept. Names the inversion from the original proposal (designing for others) to the current work (the researcher's own practice as site and subject). - Introduces PhD-Live and the wider infrastructure briefly. Sets out the thesis structure and explains the distributed autoethnographic and pedagogical approach, so the reader knows what to expect. _Conceptual argument:_ Positions the research within practice-based and autoethnographic traditions, then builds a methodological framework through speculative design, design fiction, tie-in the idea of possibility spaces to speculative design practice (thought experiments, Dunne and Raby) and live coding as a playful and performative practice. Argues that "building with and through" LLM systems is the primary mode of investigation, where each artefact is both technical object and theoretical proposition. Addresses liveness as a methodological commitment, not an aesthetic one. Names and justifies the structural decision to distribute pedagogy and autoethnographic reflection across all chapters rather than isolating them. _Conceptual argument:_ - Positions the research within practice-based and autoethnographic traditions, then builds a methodological framework through speculative design (thought experiments, Dunne and Raby), design fiction, establishing the idea of possibility spaces as speculative design practice that is being employed as foundational approach, which led to exploring live coding as a playful and performative practice that embodies liveness. Addresses liveness as a methodological commitment, not an aesthetic one. - Argues that 'building with and through' LLM systems is the primary mode of investigation, where each artefact is both technical object and theoretical proposition. - Names and justifies the structural decision to distribute pedagogy and autoethnographic reflection across all chapters rather than isolating them. _Pedagogical grounding:_ Positions the methodology in relation to critical and emancipatory pedagogy (Freire, hooks). The choice of a playful methodology is not neutral: it is a response to how learning gets flattened in institutional and commercial AI contexts. _Pedagogical grounding:_ Positions the methodology in relation to critical and emancipatory pedagogy (Freire, hooks). The choice of a playful methodology is not neutral: it is a response to how learning gets flattened in institutional and commercial AI contexts. (play as way to learn sources Montessori approach) _Autoethnographic account:_ Reflects on occupying the triple role of learner, researcher, and teacher simultaneously, and what that does to methodological choices. Through this, surfaces how the experience of teaching creative computing while researching AI's role in it produces a specific kind of knowledge that neither role alone could generate. _Autoethnographic account:_ Reflects on occupying the triple role of learner, researcher, and teacher simultaneously, and what that does to methodological choices. _Conceptual argument:_ Traces how intelligence has been defined, measured, and operationalised, and how that history has shaped the development of AI (Crawford, critical AI studies, histories of psychometrics and classification). Examines how these logics enter educational contexts through commercial AI systems, and what assumptions about agency, autonomy, and reasoning they carry. Introduces the bots here as critical responses: artefacts built to interrogate these assumptions rather than reproduce them. Addresses the political economy of commercial AI production and the power dynamics that determine how AI gets adopted within institutions and who gets to question it. _Conceptual argument:_ Traces how intelligence has been defined, measured, and operationalised, and how that history has shaped the development of AI (Crawford, Agüera y Arcas, Hayles, critical AI studies, histories of psychometrics and classification). The chatbot lineage matters here: Weizenbaum built ELIZA as a critique not a product, and that parallel to building bots that interrogate rather than optimise is worth making explicit. Examines how these logics enter educational contexts through commercial AI systems, and what assumptions about agency, autonomy, and reasoning they carry. Introduces the bots here as critical responses: artefacts built to interrogate these assumptions rather than reproduce them. Addresses the political economy of commercial AI production and the power dynamics that determine how AI gets adopted within institutions and who gets to question it. _Pedagogical grounding:_ Definitions of intelligence determine what gets optimised in educational technology. Critical pedagogy challenges the chain from measurement to design. This chapter establishes the stakes before the thesis turns to knowledge. _Pedagogical grounding:_ Definitions of intelligence determine what gets optimised in educational technology. Manning on learning otherwise and refusal of institutional categories: learning exceeds what intelligence frameworks measure. This chapter establishes the stakes before the thesis turns to knowledge. _Autoethnographic account:_ Through the experience of building tools that deliberately resist the resolve-and-optimise logic, reflects on what it means to construct alternative models of intelligence in practice. The echo chamber problem in the supervisor bot surfaces here too, from a different angle: not as a technical finding (that is chapter 4) but as a consequence of how LLM "intelligence" actually works. Through navigating institutional AI adoption as both staff and student, reflects on precarity and structural position as conditions that shape what questions feel askable. _Autoethnographic account:_ Through the experience of building tools that deliberately resist the resolve-and-optimise logic, reflects on what it means to construct alternative models of intelligence in practice. The echo chamber problem in the supervisor bot surfaces here, not as a technical finding (that is chapter 4) but as a consequence of how LLM "intelligence" actually works. The sycophancy literature (Cheng et al, Malmqvist) connects: these are structural tendencies in LLMs, not incidental failures. Through navigating institutional AI adoption as both staff and student, reflects on precarity and structural position as conditions that shape what questions feel askable.

- **process** - digital gardens, tools for thought - **process** - digital gardens, tools for thought, personal knowledge management and 'second brains' --- ### 1: introduction Frames the research questions: what AI can add to learning and knowledge-making without eroding the generative space creative work depends on, and what assumptions about intelligence and agency are embedded in commercial AI systems entering education. Introduces "After Intelligence" as a layered framing: the pun on AI, the temporal gesture past current hype, the critical interrogation of intelligence as a concept. Names the inversion from the original proposal (designing for others) to the current work (the researcher's own practice as site and subject). Introduces PhD-Live and the wider infrastructure briefly. Sets out the thesis structure and explains the distributed autoethnographic and pedagogical approach, so the reader knows what to expect. --- ### 2: methodology, a playful methodology _Conceptual argument:_ Positions the research within practice-based and autoethnographic traditions, then builds a methodological framework through speculative design, design fiction, tie-in the idea of possibility spaces to speculative design practice (thought experiments, Dunne and Raby) and live coding as a playful and performative practice. Argues that "building with and through" LLM systems is the primary mode of investigation, where each artefact is both technical object and theoretical proposition. Addresses liveness as a methodological commitment, not an aesthetic one. Names and justifies the structural decision to distribute pedagogy and autoethnographic reflection across all chapters rather than isolating them. _Pedagogical grounding:_ Positions the methodology in relation to critical and emancipatory pedagogy (Freire, hooks). The choice of a playful methodology is not neutral: it is a response to how learning gets flattened in institutional and commercial AI contexts. _Autoethnographic account:_ Reflects on occupying the triple role of learner, researcher, and teacher simultaneously, and what that does to methodological choices. Through this, surfaces how the experience of teaching creative computing while researching AI's role in it produces a specific kind of knowledge that neither role alone could generate. --- ### 3: intelligence, power and ethics _Conceptual argument:_ Traces how intelligence has been defined, measured, and operationalised, and how that history has shaped the development of AI (Crawford, critical AI studies, histories of psychometrics and classification). Examines how these logics enter educational contexts through commercial AI systems, and what assumptions about agency, autonomy, and reasoning they carry. Introduces the bots here as critical responses: artefacts built to interrogate these assumptions rather than reproduce them. Addresses the political economy of commercial AI production and the power dynamics that determine how AI gets adopted within institutions and who gets to question it. _Pedagogical grounding:_ Definitions of intelligence determine what gets optimised in educational technology. Critical pedagogy challenges the chain from measurement to design. This chapter establishes the stakes before the thesis turns to knowledge. _Autoethnographic account:_ Through the experience of building tools that deliberately resist the resolve-and-optimise logic, reflects on what it means to construct alternative models of intelligence in practice. The echo chamber problem in the supervisor bot surfaces here too, from a different angle: not as a technical finding (that is chapter 4) but as a consequence of how LLM "intelligence" actually works. Through navigating institutional AI adoption as both staff and student, reflects on precarity and structural position as conditions that shape what questions feel askable. --- ### 4: process, digital gardens, tools for thought, and knowledge infrastructure _Conceptual argument:_ Reviews digital gardens, personal knowledge management, tools for thought (Matuschik, Nielsen, Appleton), and the "second brain" tradition. Positions PhD-Live within this lineage but argues it does something these frameworks don't: it treats process visibility as a critical response to AI's tendency to collapse process into product. Introduces the local research infrastructure (supervisor bot, study companion, shared context layer) as constructed tools for thinking, not autonomous agents. Documents what they are and how they work. _Pedagogical grounding:_ Tools for thought are fundamentally claims about how learning happens. This section examines whose model of learning these tools encode, and what happens when the learner is also the builder. _Autoethnographic account:_ Through the experience of building and using PhD-Live and the local infrastructure, deeper questions about learning surface. The echo chamber problem in the supervisor bot as a finding about how correlated context collapses the space for thinking. The capability gap between local models and Claude as a finding about the sovereignty/quality tradeoff. What it means to learn in public when the infrastructure is yours. --- ### 5: knowledge, power and ethics _Conceptual argument:_ Shifts from how intelligence is defined to what counts as knowledge and who gets to produce it. Draws on tacit knowledge traditions (Schon, Polanyi, Dreyfus) to argue that the forms of knowing central to creative practice resist the explicit, extractable model that LLMs embody. Examines how institutional and commercial knowledge systems encode judgements about what is legitimate, rigorous, or valuable. Connects to the politics of knowledge within higher education: whose knowledge counts, what gets credentialled, how AI systems interact with those existing hierarchies. _Pedagogical grounding:_ This is where Freire and hooks do their deepest work. Knowledge-making as a political act, not a neutral process. The question of how AI participates in creative thinking without speaking for the researcher is fundamentally a question about knowledge and power. _Autoethnographic account:_ Through the experience of making knowledge in public via PhD-Live while navigating institutional structures that assess and credential that knowledge, reflects on the tension between open, unfinished thinking and the institutional demand for legibility. The capability gap between local models and frontier systems surfaces here as a finding about knowledge-sovereignty: who controls the infrastructure of your own thinking, and what you give up to maintain that control. --- ### 6: possibility or (im)possibility spaces _Conceptual argument:_ What did the possibility spaces offer? What happened, a moment for some reflection on the speculation? But also impossibility: the constrained, material, political reality of higher education as an institution. The tension between what can be imagined and what can be built given actual resources, time, precarity, and institutional pressure. The decision to build locally rather than rely on commercial APIs as both a political and practical choice shaped by these constraints. This is a reflective, evaluative chapter: the speculative framing meets what actually happened. _Pedagogical grounding:_ What kinds of learning become possible or impossible depending on infrastructure, institutional context, and who controls the tools? The speculative move is not abstract: it asks what learning could look like if the conditions were otherwise. _Autoethnographic account:_ Through the experience of building within real constraints (time, capacity, institutional position, the limits of local hardware), reflects on what the impossibility spaces reveal about the conditions of knowledge-making that the possibility spaces alone cannot. The gap between what was designed and what was actually buildable becomes itself a finding about the political economy of creative research. --- ### 7: conclusion Returns to the research questions. Synthesises the contributions across chapters. States the originality claim plainly: that the work treats building with and through AI as a method for investigating the conditions under which knowledge-making happens within a creative institution. Reflects on what "After Intelligence" means at the end of the project, having lived through it. Discusses limitations and future directions, including ICLC and the ongoing life of PhD-Live beyond the thesis.