Political science professors

Three parts of your work where AI is already doing real lifting, and what stays yours.

• Seminar + printed journal + library1857–1940

  Seminar + printed journal + library (pre-quantitative era)

  The defining technology of early political science was the German seminar model: a small group of graduate students reading primary texts -- Aristotle, Machiavelli, Hobbes, Locke, constitutions, statutes, diplomatic dispatches -- under a professor's direction, then writing original research. Columbia's School of Political Science, founded in 1880, operated this way. The American Political Science Review, founded 1906, was the key transmission mechanism for research. The library catalogue card, the bound journal run, and the typewriter were the principal tools. Faculty prepared lectures by hand or typewriter, duplicated documents by mimeograph for seminar reading packets, and cited sources through note cards organized in library-style boxes.

• Survey research + punch-card statistica…1948–1975

  Survey research + punch-card statistical computing (behavioral revolution)

  The behavioral revolution in political science -- roughly 1945-1970 -- transformed the discipline from a primarily legal and institutional approach into an empirical social science. The Survey Research Center at Michigan (founded 1946) and the Bureau of Applied Social Research at Columbia pioneered large-sample survey methodology. The first National Election Study was conducted in 1948. Faculty who wanted to do quantitative political science needed access to IBM punch-card machines, then early mainframe computers (the IBM 360 arrived at research universities in the late 1960s). Statistical packages like SPSS (first release 1968) and SAS (first release 1972) became the tools of the empirical political scientist. This era split the discipline: quantitative faculty at research universities gained access to expensive computing infrastructure; political theorists, comparativists working on non-democratic cases, and international relations scholars remained in the qualitative seminar tradition.

  Effect on the work

  The behavioral revolution drove a structural hierarchy in the discipline: research universities with computing infrastructure produced empirical work that dominated top journals, while teaching colleges remained qualitative. This infrastructure gap persisted until personal computers democratized statistical analysis in the 1980s.

• Xerox photocopier1960–1985

  Xerox photocopier (Xerox 914, 1960) and mimeograph for course materials

  The Xerox 914, introduced in 1960, was the first office machine to make permanent copies on ordinary paper. As prices dropped through the late 1960s, photocopiers spread to university departments and libraries, transforming course preparation. Faculty could now assemble article-length course packets from journal issues without retyping or hand-copying. The mimeograph had preceded the copier for lecture notes and syllabi since the 1920s; the Xerox replaced it for higher-quality reproduction. This era also introduced the course packet -- a stapled bundle of photocopied journal articles and book chapters -- as the standard delivery vehicle for assigned readings in political science seminars, replacing the textbook as the primary content medium at research universities.

• Personal computer + email + Stata/SPSS…1984–2000

  Personal computer + email + Stata/SPSS on the desktop

  The IBM PC (1981) and Apple Macintosh (1984) brought statistical computing to individual faculty desks for the first time, democratizing quantitative methods across the discipline. Stata (first release 1985) and the desktop version of SPSS brought regression analysis, cross-tabulation, and time-series tools to any faculty office with a PC. Email arrived at most research universities between 1988 and 1993, accelerating collaboration and manuscript exchange. Word processors -- WordPerfect dominated academic writing in the late 1980s, then Microsoft Word -- replaced the typewriter and allowed faculty to revise and reformat manuscripts without retyping. Laser printers (widely available by 1990) produced professional-quality syllabi, handouts, and manuscripts without going through the department secretary.

  Effect on the work

  The personal computer era substantially reduced the labor support infrastructure political science departments had relied on: department secretaries who typed manuscripts, maintained correspondence, and reproduced course materials. The 1980s-90s saw steady reduction in non-faculty department staff as these tasks moved to faculty computers. The flip side: faculty who mastered PC-based statistical analysis could now do empirical research that previously required a computing center with a dedicated programmer.

• Internet + JSTOR + course management sy…1995–2015

  Internet + JSTOR + course management systems (Blackboard, Moodle)

  JSTOR launched in 1995 as a digital archive of academic journals, initially focused on a small set of back-issue runs. By 2003 it had grown to cover most of the major political science journals including APSR, the American Journal of Political Science, and Comparative Political Studies -- making thirty years of literature searchable at the desktop. The internet transformed political science research by making government documents (congressional records, executive orders, treaty texts, international organization publications) freely downloadable, and by enabling access to large datasets (the Correlates of War Project, the Polity IV democracy scores) without institutional data transfers. Course management systems -- Blackboard (widely adopted 1999-2005) and later Moodle -- digitized the syllabus, assignment submission, grade book, and discussion forum. Political science simulations (Model UN, mock Congress) began incorporating online preparation tools.

• Text-as-data methods2010–2020

  Text-as-data methods (quantitative text analysis, Python/R for political science)

  The 2010s brought a new methodological wave to political science research and teaching: computational text analysis at scale. Justin Grimmer and Brandon Stewart's 2013 paper "Text as Data" laid out a framework for using machine learning on political text corpora -- congressional speeches, party platforms, newspaper editorials, UN resolutions, social media. Python and R became standard research tools at doctoral programs, with packages like tm, quanteda, and stm (Structural Topic Model) enabling analysis of datasets too large for any manual coding effort. APSA began recommending computational methods training as a doctoral competency. For political science teachers, this era meant adding a new strand to methods courses: alongside regression analysis and survey design, faculty were now expected to teach or at least introduce text-as-data tools, creating a substantial professional development demand for a cohort that had been trained before these methods existed.

  Effect on the work

  Text-as-data methods eliminated a large category of research assistant work: hand-coding of political text (scoring the ideology of congressional speeches, categorizing press releases by topic, coding news articles for political bias) shifted from labor-intensive RA projects to automated pipelines. This directly affected the labor demand for political science doctoral students doing RA work, compressing the funding model of some graduate programs.

• Large language models2020–present

  Large language models and AI research tools (Elicit, NotebookLM, ChatGPT, BERTopic)

  The arrival of GPT-3 (2020) and then GPT-4, Claude, and their successors transformed both the research toolkit and the academic integrity landscape for political science faculty. On the research side, tools like Elicit can screen thousands of papers for a systematic review in minutes; NotebookLM synthesizes working-paper corpora; BERTopic and PolitBERT enable topic modeling of congressional speech and social media at scales no human coder could match; ATLAS.ti 24 provides AI-assisted first-pass coding of qualitative interview transcripts. A 2025 study (Tornberg, Social Science Computer Review) documented that LLMs now outperform human expert coders on political social media annotation tasks -- the clearest evidence that a core research task has been substantially automated. On the teaching side, the APSA Presidential Task Force on AI (Tucker, Persily et al., 2026) documented that 61% of political science faculty had experimented with AI tools in course design by spring 2025, but also that policy brief and analytical essay assignments -- the discipline's most common assessment format -- are now among the top-10 AI-generated submission types detected by Turnitin.

  Effect on the work

  LLMs automated a significant share of the RA-level text-coding work that doctoral students had previously performed, compressing graduate research funding models and changing the nature of doctoral training in empirical methods. On the teaching side, the need to redesign AI-resistant assessments added a new professional task that consumed substantial faculty time with no corresponding reduction in other responsibilities.