Post-bacc
The National Institute of Standards and Technology (NIST) is building a library of sector-specific scenarios to facilitate AI evaluations and measurements. These scenarios are grounded in real-world AI use cases and developed with input from the broader AI community and sector stakeholders. Well-defined key elements of scenarios will ensure AI evaluations are both realistic and effective. The PREP candidate will be responsible for assisting in refining and documenting the AI scenario collection and generation process. The candidate will actively participate in NIST measurement science and be involved in human-centered research and evaluations of AI technologies.
Researching Scenarios for AI Evaluations and Metrics
- Background in any of the following or comparable fields: Computer Science, Human-Computer Interaction (HCI), Industrial/Organizational (I/O) Psychology, Cognitive Psychology, Human Factors/Engineering Psychology, Psychometrics, Economics.
- Education level: graduate student or higher (postdoc preferred).
- Strong background in research methodology.
- Competency in quantitative and/or qualitative research and data analysis.
- Knowledge/interest in human-computer interaction and human-AI interaction.
- Knowledge/interest in machine learning and AI test and evaluation.
- Ability to work both in teams and independently.
Key responsibilities will include but are not limited to:
- Developing methodology to assess various dimensions of the “goodness” of scenarios for AI evaluations, such as, but not limited to:
- Measurability of risks and benefits in AI scenarios, both at the individual and organizational levels.
- Applicability of scenarios to various evaluation types, such as evaluations to elicit negative impacts (e.g., risks), evaluations to elicit positive impacts (e.g., benefits), and evaluations of human-AI interaction.
- Operationalizing higher-level key performance indicators (KPIs) and metrics of AI scenarios into meaningful measures of risks and benefits.
- Refining and documenting the AI scenario collection and generation process for replicability and efficient scenario library development.
- Presenting results at internal meetings and occasional meetings with external stakeholders.
- Ensuring that results, protocols, and documentation have been archived or otherwise transmitted to the larger organization.