When writing program code, software developers often work in pairs—a practice that reduces errors and encourages knowledge sharing. Increasingly, AI assistants are now being used for this role.

But this shift in working practice isn’t without its drawbacks, as a new empirical study by computer scientists in Saarbrücken reveals. Developers tend to scrutinize AI-generated code less critically and they learn less from it. These findings will be presented at the 40th IEEE/ACM International Conference on Automated Software Engineering (ASE 2025) in Seoul.

When two software developers collaborate on a programming project—known in technical circles as pair programming—it tends to yield a significant improvement in the quality of the resulting software.

“Developers can often inspire one another and help avoid problematic solutions. They can also share their expertise, thus ensuring that more people in their organization are familiar with the codebase,” explains Sven Apel, professor of computer science at Saarland University.

Together with his team, Apel has examined whether this collaborative approach works equally well when one of the partners is an AI assistant. In the study, 19 students with programming experience were divided into pairs: Six worked with a human partner, while seven collaborated with an AI assistant. The methodology for measuring knowledge transfer was developed by Niklas Schneider as part of his bachelor’s thesis.

For the study, the researchers used GitHub Copilot, an AI-powered coding assistant introduced by Microsoft in 2021, which—like similar products from other companies—has now been widely adopted by software developers. These tools have significantly changed how software is written.

“It enables faster development and the generation of large volumes of code in a short time. But this also makes it easier for mistakes to creep in unnoticed, with consequences that may only surface later on,” says Apel. The team wanted to understand which aspects of human collaboration enhance programming and whether these can be replicated in human-AI pairings. Participants were tasked with developing algorithms and integrating them into a shared project environment.

“Knowledge transfer is a key part of pair programming,” Apel explains. “Developers will continuously discuss current problems and work together to find solutions. This does not involve simply asking and answering questions, it also means that the developers share effective programming strategies and volunteer their own insights.”

According to the study, such exchanges also occurred in the AI-assisted teams—but the interactions were less intense and covered a narrower range of topics.

“In many cases, the focus was solely on the code,” says Apel. “By contrast, human programmers working together were more likely to digress and engage in broader discussions and were less focused on the immediate task.”

One finding particularly surprised the research team: “The programmers who were working with an AI assistant were more likely to accept AI-generated suggestions without critical evaluation. They assumed the code would work as intended,” says Apel. “The human pairs, in contrast, were much more likely to ask critical questions and were more inclined to carefully examine each other’s contributions.”

He believes this tendency to trust AI more readily than human colleagues may extend to other domains as well, stating, “I think it has to do with a certain degree of complacency—a tendency to assume the AI’s output is probably good enough, even though we know AI assistants can also make mistakes.

Apel warns that this uncritical reliance on AI could lead to the accumulation of “technical debt,” which can be thought of as the hidden costs of the future work needed to correct these mistakes, thereby complicating the future development of the software.

For Apel, the study highlights the fact that AI assistants are not yet capable of replicating the richness of human collaboration in software development.

“They are certainly useful for simple, repetitive tasks,” says Apel. “But for more complex problems, knowledge exchange is essential—and that currently works best between humans, possibly with AI assistants as supporting tools.”

Apel emphasizes the need for further research into how humans and AI can collaborate effectively while still retaining the kind of critical eye that characterizes human collaboration.

A fermentation byproduct might help to solve two major global challenges: world hunger and the environmental impact of fast fashion. The leftover yeast from brewing beer, wine or even to make some pharmaceuticals can be repurposed to produce high-performance fibers stronger than natural fibers with significantly less environmental impact, according to a new study led by researchers at Penn State and published in the Proceedings of the National Academy of Sciences.

The yeast biomass—composed of proteins, fatty molecules called lipids and sugars—left over from alcohol and pharmaceutical production is regarded as waste, but lead author Melik Demirel, Pearce Professor of Engineering and Huck Chair in Biomimetic Materials at Penn State, said his team realized they could repurpose the material to make fibers using a previously developed process.

The researchers successfully achieved pilot-scale production of the fiber—producing more than 1,000 pounds—in a factory in Germany, with continuous and batch production for more than 100 hours per run of fiber spinning.

They also used data collected during this production for a lifecycle assessment, which assessed the needs and impact of the product from obtaining the raw fermentation byproduct through its life to disposal and its cost, and to evaluate the economic viability of the technology. The analysis predicted the cost, water use, production output, greenhouse gas emissions and more at every stage.

Ultimately, the researchers found that the commercial-scale production of the fermentation-based fiber could compete with wool and other fibers at scale but with considerably fewer resources, including far less land—even when accounting for the land needed to grow the crops used in the fermentation processes that eventually produce the yeast biomass.

“Just as hunter-gatherers domesticated sheep for wool 11,000 years ago, we’re domesticating yeast for a fiber that could shift the agricultural lens to focus far more resources to food crops,” said Demirel, who is also affiliated with the Materials Research Institute and the Institute of Energy and the Environment, both at Penn State.

“We successfully demonstrated that this material can be made cheaply—for $6 or less per kilogram, which is about 2.2 pounds, compared to wool’s $10 to $12 per kilogram—with significantly less water and land but improved performance compared to any other natural or processed fibers, while also nearly eliminating greenhouse gas emissions. The saved resources could be applied elsewhere, like repurposing land to grow food crops.”

Waste not, want not

Demirel’s team has spent over a decade developing a process to produce a fiber from proteins. Inspired by nature, the fiber is durable and free of the chemicals other fibers can leave in the environment for years.

“We can pull the proteins as an aggregate—mimicking naturally occurring protein accumulations called amyloids—from the yeast, dissolve the resulting pulp in a solution, and push that through a device called a spinneret that uses tiny spigots to make continuous fibers,” Demirel said, explaining the fibers are then washed, dried and spun into yarn that can then be woven into fabric for clothes.

He also noted that the fibers are biodegradable, meaning they would break down after disposal, unlike the millions of tons of polyester clothing discarded every year that pollutes the planet.

“The key is the solution used to dissolve the pulp. This solvent is the same one used to produce Lyocell, the fiber derived from cellulose, or wood pulp. We can recover 99.6% of the solvent used to reuse it in future production cycles.”

The idea of using proteins to make fiber is not new, according to Demirel, who pointed to Lanital as an example. The material was developed in the 1930s from milk protein, but it fell out of fashion due to low strength with the advent of polyester.

“The issue has always been performance and cost,” Demirel said, noting the mid-20th century also saw the invention of fibers made from peanut proteins and from corn proteins before cheap and stronger polyester ultimately reigned.

Freeing land from fiber to produce food

Beyond producing a quality fiber, Demirel said, the study also indicated the fiber’s potential on a commercial scale. The models rolled their pilot-scale findings into simulated scenarios of commercial production. For comparison, about 55,000 pounds of cotton are produced globally every year and just 2.2 pounds—about what it takes to make one T-shirt and one pair of jeans—requires up to 2,642 gallons of water. Raw cotton is relatively cheap, Demirel said, but the environmental cost is staggering.

“Cotton crops also use about 88 million acres, of farmable land around the world—just under 40% of that is in India, which ranks as ‘serious’ on the Global Hunger Index,” Demirel said.

“Imagine if instead of growing cotton, that land, water, resources and energy could be used to produce crops that could feed people. It’s not quite as simple as that, but this analysis demonstrated that biomanufactured fibers require significantly less land, water and other resources to produce, so it’s feasible to picture how shifting from crop-based fibers could free up a significant amount of land for food production.”

In 2024, 733 million people—about one in 12—around the world faced food insecurity, a continued trend that has led the United Nations to declare a goal of Zero Hunger to eliminate this issue by 2030. One potential solution may be to free land currently used to grow fiber crops to produce more food crops, according to Demirel.

Current production methods not only use significant resources, he said, but more than 66% of clothing produced annually in the U.S. alone ends up in landfills. Demirel’s approach offers a solution for both problems, he said.

“By leveraging biomanufacturing, we can produce sustainable, high-performance fibers that do not compete with food crops for land, water or nutrients,” Demirel said. “Adopting biomanufacturing-based protein fibers would mark a significant advancement towards a future where fiber needs are fulfilled without compromising the planet’s capacity to nourish its growing population. We can make significant strides towards achieving the Zero Hunger goal, ensuring everyone can access nutritious food while promoting sustainable development goals.”

Future of fiber

Demirel said the team plans to further investigate the viability of fermentation-based fibers at a commercial scale.

The team includes Benjamin Allen, chief technology officer, and Balijit Ghotra, Tandem Repeat Technologies, Inc., the spin-off company founded by Demirel and Allen based on this fiber production approach. The work has a patent pending, and the Penn State Office of Technology Transfer licensed the technology to Tandem Repeat Technologies. Other co-authors include Birgit Kosan, Philipp Köhler, Marcus Krieg, Christoph Kindler and Michael Sturm, all with the Thüringisches Institut für Textil- und Kunststoff-Forschung (TITK) e. V. in Germany.

“In my lab at Penn State, we demonstrated we could physically make the fiber,” Demirel said. “In this pilot production at the factory, together with Tandem and TITK, we demonstrated we could make the fiber a contender in the global fiber market. Sonachic, an online brand formed by Tandem Repeat, makes this a reality. Next, we will bring it to mass market.”

How much do undergraduate computer science students trust chatbots powered by large language models like GitHub Copilot and ChatGPT? And how should computer science educators modify their teaching based on these levels of trust?

These were the questions that a group of U.S. computer scientists set out to answer in a study that will be presented at the Koli Calling conference Nov. 11 to 16 in Finland. In the course of the study’s few weeks, researchers found that trust in generative AI tools increased in the short run for a majority of students.

But in the long run, students said they realized they needed to be competent programmers without the help of AI tools. This is because these tools often generate incorrect code or would not help students with code comprehension tasks.

The study was motivated by the dramatic change in the skills required from undergraduate computer science students since the advent of generative AI tools that can create code from scratch. The work is published on the arXiv preprint server.

“Computer science and programming is changing immensely,” said Gerald Soosairaj, one of the paper’s senior authors and an associate teaching professor in the Department of Computer Science and Engineering at the University of California San Diego.

Today, students are tempted to overly rely on chatbots to generate code and, as a result, might not learn the basics of programming, researchers said. These tools also might generate code that is incorrect or vulnerable to cybersecurity attacks. Conversely, students who refuse to use chatbots miss out on the opportunity to program faster and be more productive.

But once they graduate, computer science students will most likely use generative AI tools in their day-to-day, and need to be able to do so effectively. This means they will still need to have a solid understanding of the fundamentals of computing and how programs work, so they can evaluate the AI-generated code they will be working with, researchers said.

“We found that student trust, on average, increased as they used GitHub Copilot throughout the study. But after completing the second part of the study–a more elaborate project–students felt that using Copilot to its full extent requires a competent programmer that can complete some tasks manually,” said Soosairaj.

The study surveyed 71 junior and senior computer science students, half of whom had never used GitHub Copilot. After an 80-minute class where researchers explained how GitHub Copilot works and had students use the tool, half of the students said their trust in the tool had increased, while about 17% said it had decreased. Students then took part in a 10-day-long project where they worked on a large open-source codebase using GitHub Copilot throughout the project to add a small new functionality to the codebase.

At the end of the project, about 39% of students said their trust in Copilot had increased. But about 37% said their trust in Copilot had decreased somewhat while about 24% said it had not changed.

The results of this study have important implications for how computer science educators should approach the introduction of AI assistants in introductory and advanced courses. Researchers make a series of recommendations for computer science educators in an undergraduate setting.

• To help students calibrate their trust and expectations of AI assistants, computer science educators should provide opportunities for students to use AI programming assistants for tasks with a range of difficulty, including tasks within large codebases.
• To help students determine how much they can trust AI assistants’ output, computer science educators should ensure that students can still comprehend, modify, debug, and test code in large codebases without AI assistants.
• Computer science educators should ensure that students are aware of how AI assistants generate output via natural language processing so that students understand the AI assistants’ expected behavior.
• Computer science educators should explicitly inform and demonstrate key features of AI assistants that are useful for contributing to a large code base, such as adding files as context while using the ‘explain code’ feature and using keywords such as “/explain,” “/fix,” and “/docs” in GitHub Copilot.

“CS educators should be mindful that how we present and discuss AI assistants can impact how students perceive such assistants,” the researchers write.

Researchers plan to repeat their experiment and survey with a larger pool of 200 students this winter quarter.