It is clear that the adoption of AI writing tools is increasing in areas with low educational standards
Since the release of
[2502.09747] The Widespread Adoption of Large Language Model-Assisted Writing Across Society
https://arxiv.org/abs/2502.09747
Researchers surprised to find less-educated areas adopting AI writing tools faster - Ars Technica
https://arstechnica.com/ai/2025/03/researchers-surprised-to-find-less-educated-areas-adopting-ai-writing-tools-faster/
Researchers from Stanford University, the University of Washington, and Emory University have published a research paper titled 'Large-scale language model (LLM)-assisted writing becomes widespread throughout society.' Weixin Liang and Yaohui Zhang of Stanford University are co-first authors of the paper.
The study used a dataset that included consumer complaints submitted to the Consumer Financial Protection Bureau (687,241 cases), company press releases (537,413 cases), job postings (304.3 million cases), and press releases from international organizations (15,919 cases), as well as a statistical detection system that tracks language usage patterns. We used this data to investigate how LLM-based writing tools (AI writing tools) have been adopted in various industries from January 2022 to September 2024.
The research team employs a statistical detection system that statistically identifies whether or not an AI writing tool is being used by analyzing 'changes in word frequency and language patterns' seen before and after the release of ChatGPT. This
The research team tested the accuracy of the statistical detection system using a dataset in which the percentage of content generated by AI writing tools (0-25%) was known, and found that the statistical detection system predicted the adoption rate of AI writing tools with an error rate of less than 3.3%.
The statistical detection system analyzed the dataset and found signs of the use of AI writing tools in approximately 18% of financial consumer complaints (including 30% of all complaints from Arkansas), 24% of company press releases, up to 15% of job listings, and 14% of United Nations press releases.
In addition, when comparing the adoption rate of AI writing tools between urban and rural areas, it was 10.9% in rural areas and 18.2% in urban areas, which was significantly higher. When comparing the adoption rate of AI writing tools by educational level, it was 17.4% in areas with high educational levels and 19.9% in areas with low educational levels. The educational level was determined by the rate of bachelor's degree acquisition in each region. The adoption rate of AI writing tools was particularly high in 'urban areas with low educational levels' (21.4%).
The research team pointed out, 'The geographic and demographic patterns of adoption of AI writing tools in consumer complaints show an interesting trend that deviates from previous trends in technology adoption, which show that technology adoption is generally concentrated in urban areas, higher income brackets, and more educated groups.'
In addition, when comparing the adoption rate of AI writing tools by region for complaints submitted to the Consumer Financial Protection Bureau, Arkansas was very high at 29.2%, Missouri at 26.9%, and North Dakota at 24.8%, while West Virginia was significantly lower at 2.6%, Idaho at 3.8%, and Vermont at 4.8%. In addition, the densely populated states of California and New York had the highest adoption rates at 17.4% and 16.6%, respectively.
While the urban-rural divide is as expected, using
The research team found that similar adoption rate change patterns were observed in all areas analyzed (consumer complaints, company press releases, job postings, etc.). In particular, the change pattern was widespread: 'The adoption rate rose sharply in the 3-4 months after ChatGPT launched its service in November 2022, and then stabilized in the second half of 2023.'
Additionally, the older the company, the lower the adoption rate of AI writing tools in job listings. Companies founded before 1980 had an adoption rate of less than 5%, while companies founded after 1980 were up to three times more likely to use AI writing tools. For certain occupations, the adoption rate of AI writing tools was 10-15% higher. Companies with fewer employees were also found to be more active in using AI writing tools than companies with more employees.
Comparing the adoption rate of AI writing tools in corporate press releases by industry, science and technology companies had the highest adoption rate at 16.8%, followed by business and finance-related news at 14-15.6%, and human resources and cultural topics at 13.6-14.3%.
According to a press release from the international organization, Latin America and the Caribbean had the highest adoption rate of AI writing tools, at around 20 percent, while African countries, Asia Pacific countries, and Eastern European countries had relatively low adoption rates, ranging from 11 to 14 percent.
In response, the research team wrote, 'Although urban-rural digital divides appear to persist, our finding that lower-educated areas have higher adoption of AI writing tools in consumer complaints suggests that these tools may serve as an equalizer in consumer advocacy efforts.'
The research team also noted that the results of this survey are likely an underestimate of the actual use of AI-generated content, as advances in AI have made it more difficult to detect.
The research team claims, 'Our study shows new evidence that businesses, consumers, and international organizations are heavily relying on generative AI for communication.' At the same time, 'In sensitive areas, excessive reliance on AI can result in messages that fail to address concerns or expose information to the public that is generally unreliable. Excessive reliance on AI can also lead to public distrust of the veracity of messages sent by companies,' they said, referring to the negative effects of the spread of AI writing tools.
Related Posts:
