Writing study: Position: AI/ML Deepfake Research is Misaligned with AI Generated Non-Consensual Intimate Imagery (AIG-NCII)
← ICML 2026 oral writing corpus
Li Qiwei, Wells Lucas Santo, Sarita Schoenebeck, Eric Gilbert · Official ICML 2026 oral record
Figure, table and section ledger
Editorial contribution type: Position & research agenda (primary).
Layout extraction, not a required paper template. Counts are distinct caption labels; a multi-panel figure counts once. Section spans are estimated from heading positions, including floats and whitespace. Method and limitations.
- Canonical PDF
- 15 physical pages
- Before references / appendix
- ≈ 7.6 page-equivalents
- Main figures / tables
- 2 / 1
- Attached appendix figures / tables
- 0 / 1
- Attached appendix span
- ≈ 0.88 page-equivalents
Source-bound correction: The source has an unheaded full-page supporting Table2 on p15 after references. The explicit editorial boundary is not a printed Appendix heading; it prevents that supporting table from being miscounted as a main-text table. Physical p15 rendered and visually checked.
Section-by-section lengths and source pages
The title/abstract/front matter precede the first numbered heading. A span ends at the next detected top-level heading, not at the last sentence; end-page whitespace is included. References and statements are listed separately. These are not measurements of text-only or artwork area.
| Top-level heading | Starts on PDF page | Estimated span | Share of entire PDF |
|---|---|---|---|
| 1. Introduction | 1 | ≈ 0.57 pp. | 3.8% |
| 2. The absence of AIG-NCII in harm reduction research | 2 | ≈ 1.58 pp. | 10.53% |
| 3. Review of authenticity-based interventions | 3 | ≈ 0.91 pp. | 6.07% |
| 4. Consequences of ignoring AIG-NCII | 4 | ≈ 1.5 pp. | 10% |
| 5. Recommendations | 6 | ≈ 1.36 pp. | 9.07% |
| 6. Alternative Views | 7 | ≈ 0.75 pp. | 5% |
| 7. Conclusion | 8 | ≈ 0.25 pp. | 1.67% |
| Acknowledgments | 8 | ≈ 0.05 pp. | 0.33% |
| References | 8 | ≈ 6.53 pp. | 43.53% |
| Supporting table — editorial boundary (no printed appendix heading) | 15 | ≈ 0.88 pp. | 5.87% |
All detected figure and table locations
Attached appendix
Argument and organization
- 1 Introduction; 2 The absence of AIG-NCII in harm reduction research
States a dignity-versus-authenticity thesis and describes a selected literature analysis.
- 3 Review of authenticity-based interventions; 4 Consequences
Connects technical paradigms to their threat-model limitations.
- 5 Recommendations; 6 Alternative Views; 7 Conclusion
Offers actions, responds to objections and makes ethical partnerships a condition of research.
Problem, claim and contribution
The central distinction is viewer-centric epistemic harm versus subject-centric dignity harm, not a contest for higher deepfake detection accuracy.
Abstract and Introduction, p.1; Section 4, pp.4-5 · Main paper
Argument and section order
The literature selection and engagement categories appear before criticism of detection, provenance and watermarking, giving the position an inspectable evidence chain.
Sections 2.2-3, pp.3-4 · Main paper
Figures, captions and visual explanation
The bar chart makes the claimed research mismatch prominent, while a two-by-two text table separates synthetic/authentic status from consent-related safety without displaying intimate imagery.
Figure 1, p.3; Table 1, p.5 · Main paper
Tables, comparisons and evidence
The 39-paper analysis comes from keyword, venue and citation filters; its counts should not be generalized to every paper, and the chart juxtaposes different source populations rather than a shared denominator.
Methodology and Figure 1, p.3 · Main paper
Appendix and reproducibility
The final table lists all 39 surveyed papers, their venues and mention markers, ordered by January 2026 citation count.
Table 2, p.15 · Main paper
Clarity and writing conventions
Recommendations acknowledge imperfect defenses and require domain-expert partnerships; the objection-response format engages legal responsibility, technical tractability and institutional incentives.
Sections 5-7, pp.6-8 · Main paper
What to reuse—and what not to infer
Transferable lesson: Anchor a position in an explicit harm distinction, disclose the selection funnel behind supporting counts, and turn ethical conditions into concrete research responsibilities.
Boundary: The citation-selected landscape is not a comprehensive census, and prevalence claims from particular platforms or reports are not universal generative-AI usage estimates. Proposed interventions are recommendations, not validated harm-reduction results.
Apply this lesson using the argument and evidence worksheet.
Inspection record and sources
- Checked
- 2026-10-07
- Reading status
- Writing-focused manuscript scan + selected visuals inspected
- Main text
- Writing-oriented scan of the main argument through physical p.8, with focused physical-page text inspection on 1, 2, 3, 4, 5, 6, 7, 8. Framing, section progression, mechanism/proof explanation, displays, comparison conditions and conclusions were inspected. This is not exhaustive technical reading, proof verification or reproduction.
- Appendix / supplement
- No printed appendix heading is present. An unheaded full-page supporting Table 2 follows the bibliography on physical p.15; its organization and source-selection details were inspected. No separately linked supplement inspected.
- Visually inspected PDF pages
- 3, 5
This is a writing-focused scan of the manuscript and the stated supplementary scope: framing, organization, visual explanation, evidence presentation and wording. It is not a line-by-line proof audit, reproduction of results or exhaustive scientific review.