Appendix

Data workers

Working people who handle data materials as per the specifications of their companies and paying clients, for purposes like the training of artificial intelligence (AI) or the moderation of user-generated content in online spaces. Data workers fulfill numerous roles, ranging from labelling images to responding to tickets regarding disturbing content on social media to helping evaluate large language models (LLMs). Data work can be conducted through different business models, like business process outsourcing (BPO) and digital labour platforms (DLPs). It is important to note that data work is not a fixed category of roles. It refers to a wide variety of roles, with the possibility of new roles emerging that exhibit the characteristics of data work.

Workflow

The processes, policies, infrastructure, and people that manage the working environment in which data workers operate.

Disturbing Materials and Content/ Riskier Work

Tasks and data that feature content that can cause physical and mental health harms, or put people at risk of adverse effects, either through short-term or sustained exposure. Some examples of disturbing materials include hate speech, footage and imagery of graphic violence, content featuring sexual abuse, and online posts displaying the abuse of children.

At-Risk Workers

The workers who routinely handle disturbing materials, be it for content moderation roles or otherwise, are classified as at-risk workers. This term is not being used with intent to reduce the experiences of other categories of data workers, but to highlight that some people have to face not only precarious work engagements but also handle large volumes of content that people should not have to endure under any circumstances.

Care Mechanism

The various services and measures companies design, adopt, and manage to provide aid and support to data workers. Care mechanisms can range from providing insurance to reducing or covering operating costs involved in data work.