WORKING DRAFT · Built from the October 2026 upgrade document · Not for citation
Wellbeing & Responsibility
Figure options

Thirteen figures the thesis could use

The upgrade currently has tables only. Here are candidate figures, each drawn as a working sketch, with where it would go and the argument it makes. ★ marks the five I would add to the upgrade first.

F1 ★

Conceptual framework: three lenses, one question

Neuroliberalism governing through choice, emotion & anxiety Reflexive self pure relationship, expert systems (Giddens) Relational norms kekeluargaan · hormat · rukun · sungkan Wellbeing as a contested category produced in discourse Responsibility placed on… the individual user parents & family companies professions the state read in news (RQ1, RQ2) & in producers’ talk (RQ3)
Goes in
End of Ch3 (new §3.6 “Framework”), and as the opening slide at the upgrade viva
Argues
The three literatures meet on one question: where responsibility for wellbeing lands
Effort
Low. Redraw in PowerPoint or Figma using the house palette
F2 ★

The RQ2 corpus over time: 20 “Family & Tech” columns

Goes in
§4.5, next to (or in place of) Table 4.3
Argues
It makes the limitation in §4.7 visible: 17 of 20 columns fall after Oct 2025, and harm and clinical themes cluster in the containment stage
Data
Real: dates and titles from Table 4.3. Themes are a first-pass grouping for discussion
F3 ★

Research design: sequential, triangulated

RQ1 · Public news The Guardian · 86 so far Jan 2023 – Dec 2026 RQ2 · Family journalism Jargon, WSJ · 20 columns Jun 2023 – Dec 2026 van Dijk 4-level protocol NVivo RQ3 · Indonesia FGD → IDIs · Riliv, HIMPSI Mar – May 2027 reflexive TA TIME SPACE PERSONS converge / diverge
Goes in
§4.1, above Table 4.1
Argues
The news strands set the definitions and the interviews follow them. Triangulation (Denzin 1978) is about comparison, not verification
F4 ★

Three chatbots, one paradox

conversation: scripted → open-ended therapeutic claim: none → explicit Purpose-built Wysa, Woebot · CBT scripts General-purpose ChatGPT, Claude, Gemini Companion Replika, Character.AI furthest from therapy → most therapeutic attachment bubble size = intensity of emotional use
Goes in
§2.1 (closing the section)
Argues
News merges categories that differ in claim, constraint and provider duty, and the paradox drives the whole study
Note
Positions are conceptual, not measured. Say so in the caption
F5 ★

Body politic → body personal → body delegated

Body politic mid-1980s · the nation Body personal 1990s–2003 · the self (Sointu 2005) Body delegated? 2023–2026 · the machine this study picks up where Sointu ends
Goes in
§1.6 or §3.1, and on every poster
Argues
It places the thesis as the next chapter of a documented history, in the same newspaper
Note
“Body delegated” is a coinage to test. Keep the question mark until the findings support it
F6

Reading responsibility: four levels on one article

1 · Thematic headline + lead → macroproposition; headline/body gap 2 · Schematic summary · episode · comment: where do users, chatbots, wellbeing sit? 3 · Local coherence specify · generalise · contrast: how one case becomes a trend 4 · Style & syntax lexicon · passives · nominalisation · source types → who acts, who is quoted, what is assumed, what is unsaid
Goes in
§4.3, replacing the numbered list
Upgrade tip
Use the same layout for the worked example: put real excerpts from one article on the right of each band
F7

Responsibility map (findings template)

Goes in
Findings chapters (Ch5–6) as a summary; a blank version in Ch4 to show the analytic target
Argues
Shows how the allocation of responsibility shifts across stages and between corpora at a glance
Caution
Shading here is illustrative only. Fill it from NVivo code counts or qualitative weight after analysis
F8

What regulation reaches, and what it leaves out

Everyday, non-clinical use largely unregulated → left to users, parents & company safeguards EU AI Act assumes vulnerability can be identified in advance (Teo et al. 2025) Licensed professionals Illinois 2025 ban · HIMPSI licences (UU 23/2022) Medical devices FDA: 1,500+ narrow AI tools Wysa: FDA Breakthrough; NHS talking therapies most use happens out here
Goes in
End of §2.2
Argues
Regulation covers the device and the licensed professional. Everyday use is the space where news allocates responsibility instead
F9

How care came to rest with the family: Indonesia, 1940s → 2023

Goes in
§2.3, structured around the four movements (colonial psychiatry, silent suffering, late state, digital redistribution)
Argues
Chatbots arrive in a setting where the state came late and the family has always been the place of care
F10

A bridge or a bubble?

Therapy: a bridge therapist others, the world returns the person to relationships AI partner: a bubble relief inside the relationship, not beyond it
Goes in
§3.1 (the “bubble rather than a bridge” paragraph)
Best for
Slides and public talks. Very readable, low on detail
F11

Two selves: “Who am I?” vs “Where do I stand?”

Reflexive self Relational self “Who am I?” “Where do I stand, what do I owe?” Giddens · Illouz · Turkle wellbeing = property of a person chatbot meets an individual need Magnis-Suseno · H. Geertz · White wellbeing “inheres in the dance” chatbot avoids relational risk (sungkan, rukun) a contrast, not a stereotype (§3.3 caveat)
Goes in
End of §3.3
Caution
Keep the caption’s caveat about the New Order’s political use of “harmony”
F12

Who speaks in RQ3, and how they are reached

Researcher Riliv co-founders (2) CEO · COO · IDI each Riliv psychologists FGD (90′) → IDI each role codes in outputs HIMPSI National Policy Synergy direct invite referral only (gatekeepers) independent contact licenses
Goes in
§4.6 next to Table 4.4; also useful in the ethics application
Argues
Gatekeepers refer, but each person is contacted independently, which shows the consent safeguard visually
F13

Timeline as a Gantt chart (replaces Table 4.5)

Goes in
§4.8
Argues
Shows the overlaps (collection running while ethics is reviewed), the LPDP funding line, and the buffer year
PRODUCTION NOTES

Making these thesis-ready

  • One palette everywhere: teal = emergence / RQ1, gold = normalisation / RQ2, terracotta = containment / RQ3. Use it in figures, slides and posters.
  • Tools: conceptual diagrams in Figma, PowerPoint or draw.io. The data figures (F2, F7, F13) can be regenerated from a CSV in R/ggplot2 or Python so they update as the corpus grows.
  • Captions that argue: start each caption with the claim (“News merges three different products…”), then explain the visual.
  • Accessibility: keep text labels on every coloured element so nothing depends on colour alone. Export SVG or PDF for the thesis, PNG at 300 dpi for Word.
TYPOGRAPHY NOTES

Academic best practice for figures

  • Sans-serif in figures, serif in text. APA 7 and most journal figure guidelines ask for a sans-serif font (Arial or Helvetica) inside figures. Reading text stays serif.
  • Size. No label smaller than 8 pt once the figure is placed in the thesis (and 8–14 pt across the figure).
  • Two typefaces at most across figures, posters and slides, and use them consistently.
  • Numbers in tables use aligned (tabular) digits, right-aligned in their column.
  • Captions. APA 7 puts the figure number and title above the figure and notes below. Many UK humanities styles caption below. Follow your department’s guide and be consistent.
  • Black and white first. Shading, hatching and labels carry meaning, never colour alone.