Mathematicians and statisticians

Help write software code to analyze data more efficiently

AI code generation models are highly proficient at writing data analysis code in Python, R, SQL, and other languages, including optimization, parallelization, and implementing statistical algorithms. This task is already being substantially automated by tools like GitHub Copilot and GPT-4, with humans primarily reviewing and integrating the generated code.

BLS evidence: Mentioned that 'Some help write software code to analyze data more accurately and efficiently.'

Present findings through written reports, tables, and charts

AI can generate well-formatted reports, create publication-quality visualizations, and present findings in tables and charts following standard conventions. Tools like GPT-4 with code interpreter can produce complete analytical reports from data. Light human review ensures the presentation aligns with audience needs and organizational style, but the core task is highly automatable.

BLS evidence: The page states 'They may present written reports, tables, and charts to team members, clients, and other users.'

Analyze data using specialized statistical software

AI can write and execute code in R, Python, SAS, and other statistical software, perform standard analyses, and handle data preprocessing. Code-generation models are highly proficient at statistical programming tasks. Human oversight is needed primarily for interpreting edge cases and validating that the analysis matches the research question.

BLS evidence: The page states 'mathematicians and statisticians using specialized statistical software' for analysis and 'use statistical software to analyze data and create visualizations to aid decision making.'

Identify trends and relationships within data

AI excels at pattern recognition, correlation analysis, clustering, and identifying statistical relationships in structured data. Modern ML models can detect complex nonlinear trends that humans might miss. The remaining human value is in distinguishing meaningful patterns from spurious correlations and connecting statistical trends to causal mechanisms.

BLS evidence: The page states 'In their analyses, mathematicians and statisticians identify trends and relationships within the data.'

Conduct tests to determine data validity and account for errors

AI can implement standard validation procedures, detect outliers, run sensitivity analyses, and apply error-checking algorithms systematically. Statistical software and ML models handle routine validity testing well. Humans add value in recognizing novel error patterns, understanding when standard tests are inappropriate, and making judgment calls on borderline cases.

BLS evidence: The analysis section notes 'They also conduct tests to determine the data's validity and to account for possible errors.'

Develop mathematical or statistical models to analyze data

AI systems like Claude, GPT-4, and specialized tools can develop statistical models from data specifications, select appropriate techniques, and implement them in code. However, novel problem formulation, choosing between competing modeling approaches for ambiguous real-world situations, and validating model assumptions against domain context still benefit substantially from human judgment.

BLS evidence: The duties section explicitly states mathematicians and statisticians 'develop mathematical or statistical models to analyze data,' and the page emphasizes this as central to solving problems across all fields.

Apply mathematical theories and techniques to solve practical problems

AI can apply known mathematical techniques to well-specified problems, translate problem statements into formal representations, and execute solution procedures. The gap lies in recognizing which theoretical framework applies to a novel practical problem and adapting techniques when standard approaches don't fit, requiring human mathematical intuition and domain expertise.

BLS evidence: Listed as a primary duty: 'Apply mathematical theories and techniques to solve practical problems in business, engineering, the sciences, and other fields.'

Design surveys, experiments, or opinion polls to collect data

AI can suggest survey structures, sampling strategies, and experimental designs based on research objectives and statistical principles. However, designing effective surveys requires understanding human psychology, anticipating response biases, navigating practical constraints, and making tradeoffs that depend heavily on domain knowledge and stakeholder input.

BLS evidence: Explicitly listed in duties: 'Design surveys, experiments, or opinion polls to collect data,' with statisticians determining 'the type and size of this sample for collecting data.'

Decide what data are needed to answer specific questions or problems

AI can recommend data requirements based on statistical power calculations and modeling needs for well-defined problems. However, deciding what data are truly needed requires understanding unstated business context, anticipating downstream uses, balancing cost-benefit tradeoffs, and recognizing when proxy variables might suffice—judgment calls requiring human business acumen.

BLS evidence: First duty listed: 'Decide what data are needed to answer specific questions or problems.'

Interpret data and communicate analyses to technical and nontechnical audiences

AI can generate clear explanations of statistical concepts and draft interpretations of results for different audiences. However, effective communication requires reading the room, adapting to audience reactions in real-time, building trust around counterintuitive findings, and navigating organizational politics—skills that remain distinctly human, especially for nontechnical stakeholders.

BLS evidence: Listed as a duty: 'Interpret data and communicate analyses to technical and nontechnical audiences,' with emphasis on presenting findings and discussing limitations.