15-2021.00

Mathematicians and statisticians

34,600 workers$104,350/yr medianMaster's degree
5.8

Moderate displacement risk

Composite of 5 dimensions (higher = more displacement pressure)

AI Exposure Analysis

GPT-scored exposure9/10

This occupation is almost entirely digital, involving data analysis, mathematical modeling, and coding—all domains where AI and Large Language Models excel. While high-level theoretical research and complex interdisciplinary communication require human judgment, AI can now automate significant portions of data cleaning, statistical testing, and code generation, drastically increasing individual productivity and restructuring the workflow.

Dimension Breakdown

Technical AI Exposurepressure
9.0

How many of this occupation's tasks can current or near-term AI systems perform? Based on GPT-scored analysis of 342 BLS occupations validated against 6 academic exposure indices.

Institutional Adoption Speedpressure
10.0

How quickly will firms in this sector actually deploy AI? Accounts for regulatory burden, digital maturity, competitive pressure, union density, and organizational complexity.

Worker Adaptabilityabsorption
8.0

How well can workers in this group transition to new roles? Composite of net liquid wealth (financial buffer), skill transferability, geographic job density, and age demographics.

Demand Elasticityabsorption
8.0

When AI makes this sector's output cheaper, does demand expand enough to offset job losses? High elasticity means the Jevons Paradox may preserve or even grow employment.

AI Complementarityabsorption
7.5

Is AI primarily enhancing workers in this occupation or replacing them? Based on CFO survey data where available, estimated from task composition and job dimensionality otherwise. Jobs with more distinct task clusters (high dimensionality) tend toward augmentation via the O-Ring "focus effect" — automating some tasks lets workers concentrate on remaining ones, multiplying output quality.

6 task dimensionsHigh-dimensional

CFO survey base: 7.0 +0.5 from dimensionality

Task Composition

How this occupation's work time is distributed across 8 task categories, based on O*NET work activity data.

Information Processing22%
Coordination & Mgmt19%
Interpersonal16%
Analysis & Decision15%
Communication12%
Technical / Specialized8%
Creative / Generative4%
Physical / Manual3%

Top Work Activities

Most important work activities from O*NET, ranked by importance score (1-5).

ActivityCategoryScore
Thinking CreativelyCreative / Generative4.70
Analyzing Data or InformationAnalysis & Decision4.55
Updating and Using Relevant KnowledgeInformation Processing4.35
Processing InformationInformation Processing4.30
Working with ComputersInformation Processing4.25
Training and Teaching OthersInterpersonal4.25
Making Decisions and Solving ProblemsAnalysis & Decision4.20
Getting InformationInformation Processing4.15
Interpreting the Meaning of Information for OthersCommunication3.95
Communicating with Supervisors, Peers, or SubordinatesCommunication3.50

Methodology

This page combines Karpathy's GPT-scored technical exposure (per-occupation) with four additional dimensions inherited from the parent SOC major group: institutional adoption speed, worker adaptability, demand elasticity, and AI complementarity.

Task composition is derived from O*NET work activity data, mapped to 8 internal categories. The complementarity score is adjusted by job dimensionality (Gans & Goldfarb 2024): occupations with more distinct task clusters tend toward augmentation rather than replacement.

Net displacement risk is computed as a weighted composite: exposure (30%), adoption speed (20%), adaptability (15%), demand elasticity (15%), complementarity (20%). Pressure dimensions are normalized independently from absorption dimensions, so defensive factors can fully counterbalance exposure.