Facial blurring obscures participants’ facial features to reduce the risk of direct identification in photos or videos intended for sharing or publication.

However, applying this technique requires careful consideration of several ethical and methodological factors.

Key challenges and considerations

• Loss of analytical value: Extensive blurring can mask content into obscurity, potentially rendering visual data useless for meaningful analysis or secondary reuse.
• Risk of dehumanization: Some researchers argue that blurring faces can be perceived as dehumanizing or associated with criminality, which may undermine the dignity of the participants.
• Indirect Identification: Participants can still be recognized through context-rich cues such as jewelry, clothing, tattoos, distinctive postures, or background elements like specific buildings.
• Hidden metadata: The original file may contain hidden metadata (e.g., GPS coordinates or timestamps) that could lead to the identification of the participant.

Cropping protects participant privacy by removing specific parts of an image (e.g., a person’s face or sensitive background details). While it is a straightforward method for hiding direct identifiers, it carries significant methodological and ethical risks.

Key challenges to consider

• Loss of research value: Extensive cropping can mask content into obscurity, potentially making the remaining visual data useless for meaningful analysis or secondary reuse by other researchers.
• The risk of objectification: Ethicists warn that cropping images to exclude faces can unintentionally objectify participants by focusing solely on their bodies or specific body parts. This can conflict with visual ethics and the goal of respecting the participant’s visual voice.
• Legal & licensing restrictions: If you are using images licensed under a Creative Commons “NoDerivatives” (ND) license, cropping is strictly prohibited. This restriction even applies to minor edits (e.g., cropping images for thumbnails or social media previews).
• Insufficient protection: Cropping the main subject may not be enough to ensure full anonymity if context-rich cues (e.g., distinctive clothing, jewelry, or locations) remain visible in the uncropped areas.

Black-out bars, or black masking, is an anonymization technique where solid black shapes are placed over sensitive parts of an image (e.g., a participant’s eyes or face) to prevent direct identification. While it is a method often suggested by institutional review boards, it is increasingly criticized in visual social research for its negative impact on both ethics and data quality.

Key challenges to consider

• Associations with criminality: Research indicates that black-out bars are frequently associated with criminality and dehumanization. Using this method can unintentionally stigmatize participants, stripping away their dignity and humanity.
• Masking into obscurity: Heavy-handed masking can hide so much visual information that the image becomes analytically useless for secondary researchers. Critics argue that this results in a mockery of the value of visual data and effectively silences the participants’ visual voice.
• Conflict with visual ethics: Because this technique focuses on hiding identity through crude redaction, it often fails to respect the aesthetic and social integrity of the original research encounter.
• Insufficient protection: Masking the face does not account for indirect identification. A participant might still be recognized by peers through distinctive jewelry, tattoos, clothing, or even the image’s background context.

Reenactment involves recreating images that are central to presenting research findings. This approach allows researchers to preserve the essence and meaning of a scene while guaranteeing anonymity for the original participants. 

At the same time, reenactment requires extensive resources and is typically feasible only for a limited number of selected visuals used in publications. It is therefore difficult to scale for open data practices.  From an ethical standpoint, researchers must also ensure that recreated elements remain aligned with the original ideals and self-understandings of participants. 

These considerations equally apply to more recent approaches in which scenes are recreated not by actors, but through manual illustration or AI-generated replicas for publication or public dissemination (Kamelski & Olivos, 2025; Lobinger, 2025). 

Fabrication, also referred to as visual ethical fabrication, involves modifying images through editing tools in order to obscure identity while maintaining meaning (Markham, 2012; Tiidenberg, 2018). 

Techniques include altering colors, styles, or compositions to prevent reverse image identification (Warfield et al., 2019). The aim is twofold: to hinder recognition and to reduce the risk of images being retrieved through search engines. This concern is particularly relevant given the searchability of the web and the increasing progress of machine vision technologies.

Textual description involves replacing sensitive images with verbal descriptions to protect participant identities while maintaining research integrity

Why use it?

• It acts as a form of abstraction, omitting sensitive personal details while keeping the analytical core.
• It offers an intermediate path between full data disclosure and total redaction.

The methodological challenges

• Inherently reductive: Text cannot fully capture visual “affordances” like color, texture, spatial arrangement, gesture, or affect.
• Labor-intensive: Creating context-sensitive descriptions is time-consuming and difficult to scale for large datasets.
• Interpretation bias: Descriptions are never neutral; they are filtered through the researcher’s linguistic lens, which can distort the original meaning

Technology & legal warnings

• AI risks: Automated captioning tools can introduce bias, reproduce stereotypes, or violate privacy if data is shared with third-party AI platforms.
• Copyright still applies: Describing an image does not bypass copyright laws. Permission from the rights holder is often still required for textual reproductions.
• Informed consent: Even if participants find descriptions less invasive, they must still provide informed consent, as text can sometimes reveal enough context to allow for indirect identification.