AI-replicas as ethical practice: introducing an alternative to traditional anonymisation techniques in image-based research

Abstract

This article introduces the use of AI-replicas as an alternative to traditional anonymisation methods in image-based qualitative research. It emphasises the ethical and practical dilemmas posed by current anonymisation methods, such as distortion or loss of emotional and contextual information in images, and proposes the use of AI-replicas to preserve the integrity and authenticity of visual data while ensuring participant anonymity. The article outlines the technological foundations of generative artificial intelligence (AI) and the practical application of Stable Diffusion to generate AI-replicas for anonymisation and fictionalisation purposes. Furthermore, it discusses the potential biases present in generative AI to suggest ways to mitigate these biases through careful prompt engineering and participatory approaches. The introduced approach aims to enhance ethical practices in visual research by providing a method that ensures participant anonymity without compromising the data's qualitative richness and interpretative validity.

Keywords

image anonymisation generative AI stable diffusion research methodology qualitative research visual research ethics privacy

Get full access to this article

View all access options for this article.

References

Allen

(2015) Losing face? Photo-anonymisation and visual research integrity. Visual Studies 30(3): 295–308.

Andrew (2023a) ControlNet v1.1: A complete guide. Available at: https://stable-diffusion-art.com/controlnet/ (accessed 11 January 2024).

Andrew (2023b) How to install stable diffusion on windows (AUTOMATIC1111). Available at: https://stable-diffusion-art.com/install-windows/ (accessed 11 January 2024).

AUTOMATIC1111 (2023) Stable diffusion web UI. Available at: https://github.com/AUTOMATIC1111/stable-diffusion-webui (accessed 16 December 2023).

Banks

(2001) Visual Methods in Social Research. London: Sage Publications.

Beaumont

(2022) Laion-5B: A new era of open large-scale multi-modal datasets. Available at: https://laion.ai/blog/laion-5b/ (accessed 14 January 2024).

Berger

Luckmann

(1967) The Social Construction of Reality: A Treatise in the Sociology of Knowledge. New York: Anchor Books.

Berger

(2015) Now I see it, now I don’t: Researcher’s position and reflexivity in qualitative research. Qualitative Research 15(2): 219–234.

Betker

Brooks

Wang

, et al. (2023) DALLE3: Improving image generation with better captions. arXiv. https://cdn.openai.com/papers/dall-e-3.pdf

10.

Bie

Yang

Zhou

, et al. (2023) RenAIssance: A survey into AI text-to-image generation in the era of large model. pp. 1–25. https://arxiv.org/pdf/2309.00810.pdf

11.

Bohnsack

(2008) The interpretation of pictures and the documentary method. Forum: Qualitative Social Research 9(3): 1–17.

12.

Burgos-Thorsen

Munk

(2023) Opening alternative data imaginaries in urban studies: Unfolding COVID place attachments through Instagram photos and computational visual methods. Cities 141: 104470.

13.

Common Crawl (2023) Common crawl. Available at: https://commoncrawl.org/ (accessed 14 January 2024).

14.

CompVis (2022) Stable diffusion v1 model card. Available at: https://github.com/CompVis/stable-diffusion/blob/main/Stable_Diffusion_v1_Model_Card.md#training (accessed 14 January 2024).

15.

Douglas

(2022) Anonymizing people in images using generative adversarial networks. MSc Thesis, University of Amsterdam.

16.

Elo

Kyngäs

(2008) The qualitative content analysis process. Journal of Advanced Nursing 62(1): 107–115.

17.

Emerson

Fretz

Shaw

(1995) Writing Ethnographic Fieldnotes. Chicago: The University of Chicago Press.

18.

Gan

Greiffenhagen

Licoppe

(2020) Orchestrated openings in video calls: Getting young left-behind children to greet their migrant parents. Journal of Pragmatics 170: 364–380.

19.

Gandikota

Materzyńska

Fiotto-Kaufman

, et al. (2023) Erasing concepts from diffusion models. In: Proceedings of the IEEE International Conference on Computer Vision, Paris, France, 1–6 October 2023, pp. 2426–2436. https://doi.org/10.1109/ICCV51070.2023.00230

20.

Geertz

(1973) Deep play: Notes on the balinese cockfight. Daedalus 101(1): 1–37.

21.

Graneheim

Lundman

(2004) Qualitative content analysis in nursing research: Concepts, procedures and measures to achieve trustworthiness. Nurse Education Today 24(2): 105–112.

22.

Hannes

Parylo

(2014) Let’s play it safe: Ethical considerations from participants in a photovoice research project. International Journal of Qualitative Methods 13(1): 255–274.

23.

Heikkilä

(2023) These new tools let you see for yourself how biased AI image models are. Available at: https://www.technologyreview.com/2023/03/22/1070167/these-news-tool-let-you-see-for-yourself-how-biased-ai-image-models-are/ (accessed 13 January 2024).

24.

Hukkelås

Lindseth

(2022) DeepPrivacy2: Towards realistic full-body anonymization. Available at: https://arxiv.org/pdf/2211.09454v1.pdf (accessed 14 January 2024).

25.

Imdahl

(1996) Giotto, Arenafresken: Ikonographie – Ikonologie – Ikonik. Munich: Wilhelm Fink Verlag.

26.

Jenka (2023) AI and the American Smile: How AI misrepresents culture through a facial expression. Available at: https://medium.com/@socialcreature/ai-and-the-american-smile-76d23a0fbfaf (accessed 7 July 2023).

27.

Kozinets

(2020) Netnography: The Essential Guide to Qualitative Social Media Research (3rd ed.). London: Sage Publications.

28.

Laurier

(2014) The graphic transcript: Poaching comic book grammar for inscribing the visual, spatial and temporal aspects of action. Geography Compass 8(4): 235–248.

29.

Luhmann

(1995) Social System. Stanford, CA: Stanford University Press.

30.

MacEntee

Kendrick

Flicker

(2022) Quilted cellphilm method: A participatory visual health research method for working with marginalised and stigmatised communities. Global Public Health 17(7): 1420–1432.

31.

MacLaury

(1991) Prototypes revisited. Annual Review of Anthropology 20(1): 55–74.

32.

Madiega

and European Parliamentary Research Service (2023). Generative AI and Watermarking. Brussels: European Parliament. Available at: https://www.europeansources.info/record/generative-ai-and-watermarking/

33.

Mao

Wang

Aizawa

(2023) Guided image synthesis via initial image editing in diffusion model. In: MM 2023 – Proceedings of the 31st ACM International Conference on Multimedia, Ottawa, ON, Canada, 29 October–3 November 2023, pp. 5321–5329.

34.

Markham

(2012) Fabrication as ethical practice: Qualitative inquiry in ambiguous internet contexts. Information Communication and Society 15(3): 334–353.

35.

Maximov

Elezi

Leal-Taixe

(2020) CIAGAN: Conditional identity anonymization generative adversarial networks. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 13–19 June 2020, pp. 5446–5455. Available at: https://ieeexplore.ieee.org/document/9157102/

36.

Meng

Song

, et al. (2022) SDEdit: Guided image synthesis and editing with stochastic differential equations. In: ICLR 2022 – 10th International Conference on Learning Representations, Virtual, 25–29 Apr 2022. https://arxiv.org/abs/2108.01073

37.

Midjourney

(2024) Version. Available at: docs.midjourney.com (accessed 13 January 2024).

38.

Miller

(2015) Dear critics: Addressing concerns and justifying the benefits of photography as a research method. Forum Qualitative Sozialforschung/Forum: Qualitative Social Research 16(3): Art. 27.

39.

Mishkin

Ahmad

Brundage

, et al. (2022) DALL·E 2 preview – Risks and limitations. https://github.com/openai/dalle-2-preview/blob/main/system-card.md

40.

Moore

(2012) The politics and ethics of naming: Questioning anonymisation in (archival) research. International Journal of Social Research Methodology 15(4): 331–340.

41.

Mulhall

(2003) In the field: Notes on observation in qualitative research. Journal of Advanced Nursing 41(3): 306–313.

42.

Nicoletti

Bass

(2023) Humans are biased. Generative AI is even worse. Available at: https://www.bloomberg.com/graphics/2023-generative-ai-bias/ (accessed 30 January 2024).

43.

Nutbrown

(2011) Naked by the pool? Blurring the image? Ethical issues in the portrayal of young children in arts-based educational research. Qualitative Inquiry 17(1): 3–14.

44.

OpenAI (2020) CLIP. Available at: https://github.com/OpenAI/CLIP

45.

Niles

and OpenAI (2024) DALL-E 3 API. Available at: https://help.openai.com/en/articles/8555480-dall-e-3-api (accessed 13 January 2024).

46.

Papademas

and IVSA (2009) IVSA code of research ethics and guidelines. Visual Studies 24(3): 250–257.

47.

Phillippi

Lauderdale

(2018) A guide to field notes for qualitative research: Context and conversation. Qualitative Health Research 28(3): 381–388.

48.

Podell

English

Lacey

, et al. (2023) SDXL: Improving latent diffusion models for high-resolution image synthesis. https://arxiv.org/abs/2307.01952

49.

Polat

(2022) Facilitating a ‘virtual space’ for social change during the COVID-19 pandemic: Working with high-risk population using an arts-informed method. Visual Studies 37(1–2): 22–32.

50.

Popper

(1959) The Logic of Scientific Discovery. London: Hutchinson & Co.

51.

Prosser

(2000) The moral maze of image ethics. In: Simons

Usher

(eds) Situated Ethics in Educational Research. London: Routledge, 116–132.

52.

Radford

Kim

Hallacy

, et al. (2021) Learning transferable visual models from natural language supervision. Proceedings of Machine Learning Research 139: 8748–8763.

53.

Regulation (EU) 2016/679 (2016) Regulation (EU) 2016/679 of the European Parliament and of the Council. Available at: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32016R0679&qid=1706168522662

54.

Rombach

Blattmann

Lorenz

, et al. (2022) High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, New Orleans, LA, USA, 18–24 June 2022, pp. 10674–10685.

55.

Rose

(2022) Visual Methodologies: An Introduction to Researching with Visual Materials (5th ed.). London: Sage Publications.

56.

Sayak

Park

(2023) Fine-Tuning Stable Diffusion. Available at: https://keras.io/examples/generative/finetune_stable_diffusion/ (accessed 28 July 2024).

57.

Schütz

(1972) The Phenomenology of the Social World. London: Heinemann Educational Books.

58.

Smith

Mann

Urbas

(2018) Facial recognition. In: Urbas

Mann

Smith

(eds) Biometrics, Crime and Security. Oxon: Routledge, pp. 54–70.

59.

Smith

Miller

(2022) The ethical application of biometric facial recognition technology. AI and Society 37(1): 167–175.

60.

Spencer

(2020) The face in visual representations of children. Qualitative Research 22(2): 236–250.

61.

Spencer

(2023) Visual Research Methods in the Social Sciences: Awakening Visions (2nd ed.). Oxon: Routledge.

62.

Stability-AI (2023a) SDXL-turbo model card. Available at: https://huggingface.co/stabilityai/sdxl-turbo (accessed 13 January 2024).

63.

Stability-AI (2023b) stable-diffusion-xl-base-1.0. Available at: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0 (accessed: 11 January 2024).

64.

Stability-AI (2023c) Stable diffusion version 2. Available at: https://github.com/Stability-AI/stablediffusion (accessed 16 December 2023).

65.

Stability-AI (2023d) Stable video diffusion image-to-video model card. Available at: https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt (accessed 13 January 2024).

66.

Tiidenberg

Baym

(2017) Learn it, buy it, work it: Intensive pregnancy on Instagram. Social Media and Society 3(1): 1–13.

67.

Wahle

Ruas

Mohammad

, et al. (2023) AI Usage Cards: Responsibly Reporting AI-Generated Content. In: Proceedings of the ACM/IEEE Joint Conference on Digital Libraries, 2023, Santa Fe, NM, USA, 26–30 June 2023. Association for Computing Machinery.

68.

Wang

Burris

(1997) Photovoice: Concept, methodology, and use for participatory needs assessment. Health Education & Behavior 24(3): 369–387.

69.

Wang

Singh

, et al. (2021) IMAGINE: Image synthesis by image-guided model inversion. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Nashville, TN, USA, 20–25 June 2021, pp. 3680–3689.

70.

Ward

(2017) What are you doing on Tinder? Impression management on a matchmaking mobile app. Information Communication and Society 20(11): 1644–1659.

71.

Watt

(2015) On becoming a qualitative researcher: The value of reflexivity. The Qualitative Report 12(1): 82–101.

72.

Weber

(1949) The Methodology of the Social Sciences. New York: Free Press.

73.

Weng

(2021) What are diffusion models? Available at: https://lilianweng.github.io/posts/2021-07-11-diffusion-models/ (accessed 30 January 2024).

74.

Wiles

Coffey

Robison

, et al. (2012) Ethical regulation and visual methods: Making visual research impossible or developing good practice?. Sociological Research Online 17(1): 3–12.

75.

Yang

Zhang

Song

, et al. (2024) Diffusion models: A comprehensive survey of methods and applications. ACM Computing Surveys 56(4): 1–39.

76.

Yang

Lyu

Wang

, et al. (2022) A digital mask to safeguard patient privacy. Nature Medicine 28(9): 1883–1892.

77.

Zhang

(2023) Released checkpoints. Available at: https://huggingface.co/lllyasviel/sd-controlnet-openpose#released-checkpoints (accessed 27 January 2024).

78.

Zhang

Rao

Agrawala

(2023) Adding conditional control to text-to-image diffusion models. https://arxiv.org/abs/2302.05543