This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to papers

AI and Freelancers: Has the Inflection Point Arrived?

Qiao, Rui, Qian

2025Hawaii International Conference on System Sciences (HICSS) 2025 Proceedings5 citations
Observational labor marketInformation SystemsCausalTheoretical model
LLM / Generative AIPlatforms / gig economyWriting / contentSoftware / codingAugmentation vs. substitution
DOI: 10.24251/hicss.2025.222
Abstract

Artificial intelligence (AI) has seen tremendous improvements in capabilities recently, yet how AI impacts different labor markets remains unclear. Leveraging the launch of ChatGPT, this study aims to elucidate how AI influences freelancers across different online labor markets (OLMs). Employing the Difference-in-Differences method, we discovered two distinct scenarios following ChatGPT’s launch: 1) displacement effects in translation & localization OLM, reducing freelancers’ work volume and earnings; 2) productivity effects in web development OLM, increasing freelancers’ work volume and earnings. Theoretically, we developed a Cournot competition model to explain and identify that an inflection point exists for each occupation. Before this point, human workers benefit from AI enhancements; beyond this point, human workers would be replaced. Additional analyses across various occupations consistently demonstrate the two scenarios caused by AI. Further investigation into the progression from ChatGPT 3.5 to 4.0 reveals three evolving patterns of AI’s effects, thereby reinforcing our inflection point conjecture.

Summary

Qiao, Rui, and Qian use difference-in-differences with propensity score matching on worker-level transaction data from a major online labor platform to study how ChatGPT's launch affected freelancers across 11 different occupational markets from May 2022 to October 2023.

Main Finding

ChatGPT's launch caused a 9.0% decrease in job volume and 29.7% decrease in earnings for translation & localization freelancers (displacement effects), but a 6.4% increase in job volume and 66.5% increase in earnings for web development freelancers (productivity effects)

Primary Datasets

Proprietary worker-level transaction data from unnamed online labor platform (6,743 construction design workers, 7,582 translation workers, 15,000 web development workers)

Secondary Datasets

Felten et al. (2021)

AI-focused

Google Trends

Other

AI Occupational Exposure Index (AIOE) from Felten et al. (2023); Google Search Volume Index

Key Methods
Difference-in-differences with propensity score matching comparing translation & localization and web development markets (treated) to construction design market (control) around ChatGPT's November 2022 launch; complemented by Cournot competition theoretical model
Sample Period
2022-2023
Geographic Coverage
Global (online labor platform with workers from multiple countries including US and international)
Sample Size
29,325 workers (after removing inactive workers) across 11 occupational markets observed monthly over 16 months, yielding 350,752 worker-month observations
Level of Analysis
Individual, Occupation
Occupation Classification
Platform-specific occupation/market categories (translation & localization, web development, construction design, plus 8 additional markets)
Industry Classification
None
Notes
ICIS 2024 Proceedings [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection. [Claude classification]: Paper introduces "inflection point conjecture" - theoretical model showing AI can benefit workers before crossing occupation-specific threshold but displaces them after. Analysis of ChatGPT 4.0 upgrade reveals three patterns (continuous displacement, productivity-to-displacement transition, continuous productivity) supporting irreversibility of displacement. Uses both AIOE index and Google Search Volume to validate treatment/control group selection.