Chris J. Vargo is an associate professor at the University of Colorado Boulder and a computational social scientist who studies how media, platforms, advertising, and digital information systems shape public attention and behavior. His research uses computational methods—including machine learning, digital trace data, and network analysis—to examine agenda-setting, social media dynamics, advertising, and political communication.
Current Research
How AI and digital platforms shape attention, advertising, and public life.
I study information flows across search, social media, and advertising systems. Current projects use digital trace data and machine learning to examine agenda-setting, platform behavior, consumer journeys, and the social consequences of AI.
HICSS-58, 2025.
Current Focus
Gambling advertising, gambling discourse, and media effects
Computational content analysis and machine learning for media research
Agenda-setting, public attention, and intermedia influence
Advertising analytics, digital trace data, and platform behavior
The live extension of our Policy & Internet paper, rebuilt as a working
instrument: a decade of gambling ad spend, Google searches for gambling help in every
state back to 2004, the actual ads people were shown, and what a thousand views cost
then versus now.
Journalists — is attention getting more expensive? Who pays more, who pays less? Every chart carries its receipts.
Researchers — state-by-month help-seeking and spend series, with methods and exclusions stated on the page.
Regulators & advocates — who is advertising, where the burden concentrates, and what the platforms refuse to publish.
006 TABS022 YRS OF SEARCH DATA557 ADS CENSUSED001 HEATMAP
Journal of Marketing Analytics · author-accepted manuscript · 2026
LLM assistants increasingly mediate product discovery. Using de-identified passive behavioral logs linked to a post-trace survey, we measure how LLM exposure compresses and reorders consumer journeys — and show why journey analytics must treat conversational discovery as a first-class touchpoint. In a one-month window, LLM-exposed journeys show substantially higher observable LLM and review/UGC share in pre-intent discovery, while unadjusted compression and time-to-intent differences shrink after adjustment.
American Academy of Advertising 2027 · author-accepted manuscript · Las Vegas
Five open-weight LLMs recognize ad content at 91.6% accuracy, yet stronger evidence fails to reliably move their judgments in any of 12 ad families — sham edits sway them as readily as real upgrades. Fluency without diagnostic sensitivity is the appearance of judgment, not judgment.