The invisible majority: linking a national researcher registry to OpenAlex reveals Peru's uncertified artificial intelligence workforce

Governments need to know who is doing the work of science to plan for the future. In many Latin American countries, they rely on official lists, or registries, that certify researchers. These lists act as a formal record of who counts as a scientist, often determining who gets funding or professional recognition. The assumption behind these lists is that they capture the people actually doing the research. However, science moves fast, especially in fields like artificial intelligence, where new experts emerge constantly from universities and industries. If the official lists miss the newest and most active workers, the government's picture of its scientific strength becomes incomplete, potentially leading to poor decisions about where to invest in training and resources.

A team of researchers from the Universidad Nacional del Altiplano in Peru decided to test this assumption. They wanted to see if Peru's national registry of scientists, known as RENACYT, truly reflected the country's workforce in artificial intelligence. To do this, they did not rely on surveys or estimates. Instead, they built a bridge between two massive, open databases. On one side was the official government list of certified researchers. On the other was OpenAlex, a global, free-to-use catalog of scientific papers and the people who wrote them. By connecting these two sources, the team could compare the official list against the actual output of researchers publishing papers on artificial intelligence and computer vision.

The results revealed a significant gap between the official record and reality. The researchers found that the government's registry captures only a fraction of the people producing artificial intelligence work in Peru. While the registry lists over 11,000 certified researchers, only about 856 of them are working primarily in artificial intelligence. When the team looked at the broader group of people publishing papers with Peruvian connections in this field, they found nearly 3,500 distinct authors. Of these, only about one in four held an official certification. This means that the vast majority of the people doing the work are invisible to the official statistics. The certified group represents a small core, while the uncertified majority, which includes students, recent graduates, and industry practitioners, makes up the bulk of the workforce.

This invisibility changes how we understand the growth of the field. The number of papers coming from Peru in artificial intelligence has grown fivefold over the last decade, yet the official registry has not kept pace with this surge. The new authors entering the field each year are rarely the ones getting certified immediately. Consequently, any government plan based solely on the registry would be looking at only a quarter of the actual talent pool. The study suggests that for emerging fields, these official lists should be viewed as a snapshot of a stable core rather than a complete census of everyone contributing to the science.

The study also uncovered a surprising story about where these researchers live. For decades, scientific work in Peru has been heavily concentrated in the capital city, Lima. The official registry confirms this, showing that more than half of all certified researchers live in the capital. However, when the researchers looked specifically at artificial intelligence, the map shifted. While Lima still has the largest total number of certified AI experts, other regions are actually more dense with talent. When measured by the number of researchers per person living in the area, cities in the south like Arequipa, Tacna, and Puno have more certified AI experts per capita than Lima does. In fact, the Universidad Nacional del Altiplano, located in the high-altitude region of Puno, ranks as the third most productive institution in the country for certified AI researchers, ahead of some of the nation's most famous universities.

This finding challenges the idea that the provinces are lagging behind in performance. The data shows that researchers outside the capital produce just as many papers per person as those in Lima. The difference is simply in the total number of people. The provinces are not underperforming; they are just smaller in scale. This suggests that the country's scientific strength is more spread out than the official numbers suggest, with strong pockets of expertise in the south that are ready to be supported.

The researchers also examined how the official certification levels relate to actual productivity. The registry assigns researchers to different tiers based on their career achievements, with higher tiers indicating more seniority and prestige. One might expect that those with the highest certifications would be the most productive in artificial intelligence. The data, however, shows a very weak connection. A researcher's official rank does not strongly predict how many papers they write in this specific field. Many of the most active writers in artificial intelligence are in the lower tiers of the registry, often because they are early in their careers. This indicates that the general certification system, which looks at a scientist's entire career across all topics, is not a precise tool for measuring expertise in a fast-moving, specialized area like artificial intelligence.

Finally, the study highlighted a gender gap that widens within the field. Women make up about one-third of the total certified researchers in Peru's registry. However, in the subset of artificial intelligence researchers, that number drops significantly. The proportion of women falls to less than one-fifth of the AI group, and it drops even further among the highest-certified experts in the field. This pattern suggests that the barriers for women in artificial intelligence are not just about entering the field, but also about progressing to the highest levels of recognition within it.

By linking the government's official records with the global record of published papers, this study provides a clearer, more honest picture of Peru's scientific workforce. It shows that the country has a much larger and more geographically diverse artificial intelligence community than the official lists admit. The findings do not criticize the registry itself, which remains a vital tool for policy, but they do suggest that for fast-growing fields, these lists must be checked against open data to ensure that the "invisible majority" is not overlooked. The work demonstrates that with open data and careful matching, it is possible to audit these systems and reveal the true shape of a nation's scientific potential.