As an infrastructure, Playlab is gender-neutral. As an organization, Playlab collaborates closely with educators to ensure that their voices and contexts are represented in the apps they create. Guardrails have been implemented in the infrastructure to facilitate moderation and prevent the development of inflammatory or biased applications.

Playlab has integrated voice functionality into its platform, allowing users to have the text read out loud to them, which supports individuals with vision impairments. Playlab also plans to implement technology that will enable users to interact with the system using voice commands. Moreover, Playlab serves as an infrastructure that facilitates the creation of customizable applications. This means that developers can build apps tailored to the specific needs of different user groups. We depend on community members to identify and communicate the unique requirements and circumstances of the communities they serve.

Design & Functionality: Playlab is a collaboration between educators and learners, not just engineers. We focus on catering to educational needs rather than following generic workflows and strive to accommodate various teaching methods rather than enforcing a singular approach. Our "Show Our Work" principle is evident in in-product disclosures, options for creators to review raw logs of AI choices and tool calls, as well as transparent references and resources used in interactions.

User Interface:

• Representation in feedback: Every app allows users to manually flag outputs for bias, appropriateness, and hallucination.
• Controls: We are developing content moderation that can be customized by educators/administrators within safe parameters, tailored to their students' age, context, and norms.
• Evaluation feature: Our evaluation tools empower users to define success criteria and app goals, aiding them in identifying outcomes and tailoring designs for their specific audience and context.

Business Model:

• Nonprofit structure: As a tech nonprofit, our focus is on impact and safety rather than profits. We aim to create public infrastructure.
• Open-source commitment: We prioritize open-source AI to benefit the broader field with responsible practices.
• Community-shaped: We engage a community of educators to test features and share insights, hosting in-person sessions to understand what students and educators prioritize in terms of safety.

Power Dynamics: We prioritize our responsibility to students, educators, and the future of education rather than merely avoiding blame. We establish feedback loops where the concerns of educators and students directly influence product priorities, guidelines, and policies.

Mitigating Bias:

• Bias and alignment guidance are given to AI models in all Playlab apps, in addition to their base guidelines.
• Automated moderation is implemented for all prompts and model outputs.
• Red teaming (adversarial testing) is conducted during product development.
• There is a commitment to actively identifying bias and openly sharing limitations when mistakes are made.

Transparency:

• The product includes clear disclosures about how AI functions, potential issues that may arise, and known limitations.
• A user-facing flagging system allows for identification of bias, appropriateness, and hallucination issues.
• Professional learning is provided to ensure the community comprehends the risks and can make well-informed decisions.

Accountability:

• A Trust Council serves as a standing forum where new features undergo review for trust and safety, with clear plans and test owners determined.
• The Product Requirements Document (PRD) process mandates documentation of potential harms, biases, and mitigations before features are prioritized.
• Responsible piloting with a smaller subset of users is conducted through co-design before a broader release.