Open expert roles.

Job Description

Bring senior decision science expertise to AI evaluation projects focused on decision models, uncertainty, judgment biases, and choice under risk. Review model responses and worked examples for confounded evidence, overgeneralized findings, and invalid measurement assumptions. Assess whether the reasoning meets professional standards for the stated task. Write reference answers and scoring criteria grounded in decision models, uncertainty, judgment biases, and choice under risk. Explain consequential errors, evidence gaps, and the reasoning behind your corrections. This is remote, project-based contract work. Scope, hours, and assignments depend on project matching; the listed hourly range applies to matched work.

Why Apply

• Expert-level compensationEarn $160 - $210/hr for work that values your professional judgment.
• Flexible remote workContribute asynchronously around your existing professional commitments.
• Meaningful model impactHelp advanced AI systems reason more accurately in psychology & behavioral science.
• Focused projectsAccept only the opportunities that fit your experience and availability.

Responsibilities

• Evaluate advanced domain tasks and AI-generated responses for factual accuracy and professional judgment.
• Apply current standards, methods, and risk considerations from psychology & behavioral science.
• Document errors, edge cases, and the reasoning behind an ideal response.
• Collaborate asynchronously with Alpheva's research and evaluation teams.

Required Qualifications

• 7+ years of substantive experience in decision science research or applied decision analysis.
• Demonstrated senior-level judgment in decision models, uncertainty, judgment biases, and choice under risk, with substantial responsibility for reviewing or delivering this work.
• Ability to write precise reference answers, assess competing approaches, and justify corrections with traceable evidence.

Company Description

Alpheva AI builds the expert data, evaluations, and real-world environments that help frontier AI systems become more capable and reliable. We partner with experienced professionals whose judgment can teach models how high-quality work is actually done.