皮尤：AI暴露度越高的职业，男性就业增长快于女性

一份按性别拆开的就业账本

皮尤研究中心这次没有继续追问“AI会不会让工作消失”，而是换了一个角度：在同一个职业内部，男性和女性的就业人数变化是否同步。分析使用美国当前人口调查（CPS）2017—2019年与2023年1月至2026年6月的基本月度数据，中间2020—2022年因疫情重塑就业而被排除。研究者为480个职业建立0到3.5分的AI暴露度评分，依据是美国劳工部O*NET对任务能否由生成式AI完成或协助、以及该任务对职业重要程度的描述。

结果呈现出一条清晰的斜率：职业的AI暴露度每高1分，男性就业人数增长6.3%，女性只增长0.2%。从暴露度最低到最高，男性就业增长约多出24%，女性则基本持平。在暴露度最低的一端，模式反过来——女性就业增长快于男性。

差距集中在没有四年制学位的人身上

按教育分层后，图景更具体。高中及以下学历者中，AI暴露度每高1分，男性就业增长8.8%，女性下降9.5%；受过部分大学教育者差距缩小但仍显著；拥有学士及以上学位者，男女就业变化都不显著。

作者提醒，不能据此认为高暴露职业里男性占多数。没有大学学位的男性其实更多集中在卡车驾驶、建筑劳务、货运等低暴露职业，因此“高暴露职业中男性增长更快”只适用于这部分男性中的较小比例。相反，没有四年制学位的女性更多从事秘书、前台、客服等处于暴露度中上段的职业，聊天机器人和自动客服在这类工作中的介入空间更大。

相关不等于因果

皮尤明确写道，AI暴露度与就业增长之间的关系未必是直接的。这项分析比较的是生成式AI广泛可用前后两个时段，但同一时期还有其他变化可能影响就业，例如行业周期、地区差异、岗位结构与企业用工策略。研究也没有比较不同职业之间的差异，作者特意声明，不认为某种职业天然属于男性或女性。

因此，更稳妥的读法是：在AI暴露度较高的职业内部，男女就业人数的变化出现了分化，且分化与教育程度相关；至于AI在其中起了多大作用、通过什么机制起作用，这份分析没有给出因果答案。

为什么值得放进心理学的视野

这不是一项心理学实验，但它触及的工作处境与心理生活直接相关：职业中任务被自动化或重组，会改变人对自身技能是否仍有价值的判断，也会影响工作安全感与职业认同。性别与教育的交叉差异意味着，同一波技术变化对不同群体的意义并不相同——对部分高暴露职业中的女性而言，变化可能表现为岗位数量收缩，而非任务被辅助。

这些心理层面的推论需要另外的研究来检验，皮尤的数据只到就业人数这一层。把就业统计当作情绪或身份感受的证据，会越过材料本身。

Artificial intelligence is transforming U.S. workplaces. AI chatbots like ChatGPT and Claude are bringing new efficiencies to certain work tasks and automating others. These technologies are still relatively new, and it’s too early to tell if they will lead to job loss or job growth in the long run. But there are early signs of a gender gap: Employment growth has been slower for women than for men in occupations with more exposure to AI.

We measured employment growth for men and for women in occupations with different levels of exposure to AI, based on government data. We compared the years before generative AI became widely available (2017-19) to the years since (2023-26).

Employment growth is the percent change in the number of people employed in an occupation. Within each occupation, we compared growth for men to growth for women. We did not compare occupations to each other, and we are not suggesting some occupations are inherently “men’s work” or “women’s work.”

AI exposure measures how much AI can do or help with the tasks required in a particular occupation.

About this research

Much of the public conversation about artificial intelligence (AI) and work has focused on whether jobs will disappear. But employment in occupations that are the most exposed to AI has grown, not shrunk. This Pew Research Center analysis asks a different question: How has job growth differed by gender and education?

Why did we do this?

Pew Research Center does high-quality research to inform the public, journalists and decision-makers. This research builds on our previous work on the impact of AI on jobs and public attitudes toward AI at work.

How did we do this?

This analysis used Current Population Survey (CPS) basic monthly files from 2017 through 2019 and January 2023 through June 2026. The years 2020 through 2022 were excluded because the COVID-19 pandemic reshaped employment for reasons unrelated to AI.

We measured employment using the CPS final weight.

Employment growth refers to the change in the number, not the share, of people employed. Within each occupation, we compared how the number of employed men changed to how the number of employed women changed. We did not compare occupations to each other, and we are not suggesting that any occupation is more suited to one sex.

The gender gap is the difference between men’s and women’s employment growth within each occupation. We then tested whether that gap is bigger in occupations with more AI exposure.

We built an exposure score for each occupation by looking at the tasks involved in 480 occupations, whether each task can be performed or assisted by generative AI, and how important the task is to the occupation. The scores are based on data from the Occupational Information Network (O*NET), a U.S. Department of Labor database that describes the tasks people do at work.

Scores range from zero to 3.5. A score of zero does not mean an occupation has no exposure to AI, only that it is the least exposed occupation in our data. Higher scores mean generative AI can do or assist with a larger share of the occupation’s most important tasks.

Read the technical appendix for definitions and further details.

From 2017-19 to 2023-26, men saw substantially more job growth than women in occupations with higher AI exposure.

Different occupations have different levels of AI exposure. In construction labor, AI can do very little of the core work, whereas in accounting, it can meaningfully help with much of the work. Occupations with some level of exposure to AI tend to include tasks that involve reading, writing or working with information.

This analysis finds that for each 1-point increase in an occupation’s AI exposure score, men’s employment grew by 6.3%, while women’s employment barely changed (0.2%).

Because men’s employment grew steadily with each additional point of AI exposure while women’s barely changed, the gender gap in employment growth widened step by step from the least to the most exposed occupations. Men’s employment grew about 24% more in the most AI-exposed occupations than in the least exposed, while women’s employment growth was about the same across the scale.

At the low end of the AI exposure scale, the pattern reverses: In the least AI-exposed occupations, women’s employment grew faster than men’s.

It’s important to note that the relationship between AI exposure and job growth is not necessarily direct. This analysis compares employment before and after generative AI became widely available, but other changes during this period may have also affected employment.

Related: The Gender Gap in AI

Measuring AI exposure

We scored 480 occupations on how exposed each is to generative AI, using ratings from the U.S. Department of Labor’s Occupational Information Network (O*NET) and the Current Population Survey (CPS).

Each occupation received an exposure score ranging from zero to 3.5. A score of zero does not mean an occupation has no exposure to AI. Instead, it means it’s the least exposed occupation in our data.

Higher scores mean generative AI can do or assist with a larger share of the occupation’s most important tasks.

Occupations with low AI exposure include construction laborers, carpenters, home health aides and retail sales workers. Occupations with high AI exposure include customer service representatives, accountants, software developers and financial managers.

Differences by gender and education

The gender gap in AI-exposed job growth varies widely by education. As AI exposure increased, the gap between men’s and women’s employment widened for workers without a bachelor’s degree, but not for those with one.

Among workers with a high school diploma or less, men’s employment grew by 8.8% for each 1-point increase in AI exposure, while women’s fell by 9.5%.

For workers with some college education, the gap was smaller but still significant.

For workers with a bachelor’s degree or higher, neither men’s nor women’s employment changed significantly as AI exposure increased.

To be sure, men without a college degree are not heavily represented in occupations with high levels of AI exposure. Instead, they tend to work in occupations with low AI exposure, such as truck driving, construction labor and moving freight. So the job growth we found for men without a degree in more AI-exposed occupations applies to a relatively small share of these men.

Unlike men without a four-year college degree, women with this level of education are often working in occupations like secretaries, receptionists and customer service representatives. These roles are closer to the middle to upper part of the AI exposure scale, so tools like chatbots and automated support have a larger role in their work.