Seven Out of Ten Koreans Say "Thorough Verification Needed for Public Sector AI Adoption"

PMAI Surveyed 1,000 Generative AI Users
63.8% of Respondents Say "Gender Bias Checks in AI Are Necessary"
Pre-screening Desired Especially in Areas Related to Physical Well-being and Safety

Seven out of ten Koreans believe that when public institutions adopt artificial intelligence (AI), it is important to conduct thorough verification—even if it takes time.

Research and data intelligence company PMI announced on July 5 that it conducted a survey on “perceptions of AI fairness and gender bias” among 1,000 people aged 20 to 59 who have experience using generative AI. Gender bias in AI refers to the phenomenon in which AI repeatedly produces outcomes that are advantageous or disadvantageous to a particular gender, due to biased data or algorithm design. The survey was conducted online from June 24 to 25.

The results showed that the majority of respondents called for a cautious approach to introducing AI in the public sector and addressing gender bias issues. Specifically, 68.1% said that public institutions should adopt AI only after sufficient verification, even if it takes time. This figure is more than three times higher than the 20.7% who said that quick adoption is important for convenience and efficiency.

By gender, 72.6% of women prioritized verification, compared to 63.6% of men—a difference of 9 percentage points. The tendency toward caution increased with age: 75.4% of those in their 50s said verification should come first, compared to 60.6% of those in their 20s.

The survey also found that as people gain more experience with AI, their expectations for efficiency rise. Among so-called “heavy users” who use AI for work every day, 35.8% were open to rapid adoption—about twice as high as the 17.4% among “light users.”

In addition, 63.8% of respondents said that Korea needs its own criteria for examining gender bias in AI. The Korean-style criteria would not simply adopt overseas regulatory and evaluation frameworks, but would instead reflect the Korean language, perceptions of gender roles, and labor market practices unique to Korean society, in order to judge the presence of bias.

Both men (62.0%) and women (65.6%) showed agreement rates over 60% on the need for Korea-specific standards for checking gender bias in AI. By age group, agreement was highest among those in their 50s (72.8%) and lowest among those in their 30s (57.0%), indicating some generational differences.

The areas where respondents said prior government review is necessary before AI adoption were “crime prevention and public safety” (50.0% for men, 47.5% for women) and “medical diagnosis” (44.0% for both men and women), both ranking at the top regardless of gender. In both areas, AI errors are directly linked to personal physical safety.

The most important factor for ensuring AI fairness was “the ability to explain the basis for AI’s decisions” (21.6%), followed by “a procedure for human final review” (17.6%). This indicates a demand for both transparent disclosure of AI’s decision-making grounds and a double safety mechanism where a human checks the AI’s decisions again.

Minhee Cho, CEO of PMI, said, “What the public expects is not the speed of technology adoption itself, but trustworthy verification procedures and explainability that reflect the context of Korean society,” adding, “Rather than blanket regulation, it will be necessary to develop carefully designed, step-by-step policies.”

This content was produced with the assistance of AI translation services.