Skip to main content

Scientific Pioneers Break Down Walls to Address Pressing Societal Issues at AAAS Annual Meeting

AAAS Annual Meeting Science briefing
The first-ever Science family journals editor trends briefing at the AAAS Annual Meeting drew a full contingent of attendees to hear about research on subjects including data and AI. | Robb Cohen Photography & Video

The briefing room grew crowded – it was standing-room only – as the future of science was on display in the first-ever Science family journals editor trends briefing at the AAAS Annual Meeting. 

Four panelists – each nominated by a journal editor for their pioneering work addressing society’s most pressing problems – talked about their research.  

Francesca Dominici, professor of Biostatistics at the Harvard T.H. Chan School of Public Health and director of the Harvard Data Science Initiative, uses data to help predict and reduce negative climate change-related health outcomes from air pollution, including from wildfires. 

Alessandro Roncone, assistant professor of computer science, University of Colorado Boulder, is experimenting with improving robots in various ways – including their sense of touch – so bots can better integrate into human settings.  

Melanie Mitchell, Professor at the Santa Fe Institute, is uncovering where AI has notable gaps in cognition compared to humans – and why.  

The final panelist, Brian Donovan, senior research scientist at BSCS Science Learning, discussed innovative ideas for reforming STEM education to reduce race and gender biases, with related work published in Science. 

The four panelists elaborated on their unique methods. But they also talked about common struggles. 

Across all fields, each one is challenged by lack of access to critical, often politically sensitive data, difficulty obtaining funding for work that is ever more interdisciplinary in nature, and routinely finding themselves a target of politically motivated attacks related to their chosen field. 

One reporter asked what advice the panelists would give to scientists whose work isn’t politicized now – but might be one day.  

“I don’t think we should avoid going after a scientific area because we’re worried we might be criticized,” said Dominici. “I’ve been a target for a very long time, and … I’ve tried to be resilient to the attack, to continue to work with the highest possible scientific rigor, and to continue to publish.” 

“Science in general is a consensus system – an inquiry process based on facts and data,” added Roncone. “The best way to counteract these attacks is to approach them as a community and have a consistent, unified voice. That doesn’t mean that scientists always need to unify behind a specific answer, but they should all get behind a constructive conversation around the problem.” 

“It’s very important to have great colleagues and friends,” said Donovan.  “It's a lot easier to take the blow of criticism – even when it feels unwarranted – when you have good friends and collaborators that you can talk to and strategize with before you communicate or address those issues.” 

Another common thread across this group’s pioneering work is the use of AI. 

Leading the charge, in many respects, Mitchell’s work examines how we can know how smart AI systems truly are. While there have claims suggesting AI systems are quite intelligent, what scientists really want to know when they ask related questions is whether these systems can form the kinds of abstract representations that enable humans to reason when presented with new scenarios. 

“There have been many claims of such abilities in today's generative AI systems,” Mitchell said, “but such claims can fall prey to pitfalls of evaluation.” As one example, Mitchell referenced standardized tests, like the SAT or the bar exam. ChatGPT has excelled on such examinations, she noted, but drawing conclusions of intelligence from these results is not possible because an increasing number of the related questions are included in the system's training data. The systems have “seen” it before.

To test the human-like abstraction capabilities of an AI system, a better approach, Mitchell explained, would be to use the experimental methods from developmental psychologists and animal cognition researchers. This is work she and her team is pursuing. 

“What we and others are seeing,” she said, “is that while these systems exhibit strong performances on tasks that are likely similar to content they've seen in their training data, they fail when the same tasks involve content unlikely to be similar to training data.”

“We still have a ways to go before our AI systems exhibit the same level of general and robust intelligence as humans,” said Mitchell. 

In reply to a question about fears surrounding artificial intelligence, robotics and the future of work, Roncone had a positive message to share.  

“There’s something called Moravec’s paradox,” he said. “Things that you think would be easy for robots to do are hard, and things that you think would be hard to do are easy.” 

Robots will do work that increases our bandwidth to take on more meaningful tasks, Roncone said. He gave the example of organic chemists all over the country eager for a robot-led approach his team is developing to reduce the number of monotonous and dangerous tasks human chemists have to do. 

“This is a very specific problem, but also a problem that chemists would like us to fix,” Roncone said. “We talked to a number of companies and everybody's very excited about the potential for these solutions to come to fruition.” 

Dominici talked about finding solutions to some of society’s most intractable problems by innovative uses of large datasets. In particular, her work is focused on using data on threats like air pollution to build a “climate smart” public health system, and harnessing data science to prevent gun violence. She is also focused on rigorously analyzing unveiling the digital carbon footprint – and thus the environmental impact – of data centers.

Donovan presented new work from the journal Science, in two Policy Forums. One – an analysis of popular U.S. high school biology textbooks – revealed how the books’ approach to genetics leads to essentialist ideas about sex and gender among adolescents. “Essentialism” is the assumption that categories of living things have underlying “essences.” 

The second Policy Forum reported on cluster-randomized trials in several U.S. states that showed that teaching students about genetics via an approach Donovan calls “humane genomics education” can reduce racial bias. Between December 2019 and May 2022, Donovan and his colleagues recruited 15 teachers and 1,063 biology students from six U.S. states for their analysis. 

As the briefing wrapped up, each panelist offered advice for young people who would like to get involved in pioneering fields, like their own. 

"This is an interesting time for people who are getting into AI,” said Mitchell. “A lot of the narrative is that most of the problems have now been solved, which is completely false. … And I think there are a lot of interesting niches in AI that don't necessarily involve large language models – an important thing to remind people.” 

“My advice is that it's never too early to get started with robotics,” said Roncone. “We have high schoolers that do research in my lab on a regular basis. They graduate, then they go out and do great work. Robotics is unique in that you can come to it from a hardware perspective.” 

“We talked here today a lot about artificial intelligence and generative AI, but I would also invite students to get into the field of data science,” Dominici said. “This entails taking a massive amount of information and data available everywhere – for example, satellite images of deforestation and droughts. We can combat the climate crisis just by ingesting this data, and then using it to be able to predict where the next problem is going to occur.” 

Perhaps one of the greatest joys of the event was that after it concluded, the panelists looked at each other, and one said, “I learned so much from you.” It was clear the sentiment was shared panelist-wide. 

What future awaits when innovators working on society’s biggest problems – no matter how disparate –come together and build their courage and connection? Stay tuned. The future is full of possibilities.

Authors

Meagan Phelan

Communications Director, Science Family of Journals

Matthew Wright

Related Focus Areas

Related Scientific Disciplines