While the term ‘social media influencer’ may bring to mind Instagram accounts offering hair growth gummies for a shilling or cute outfits, the truth is that influencers influence all sorts of things. Including research trends in artificial intelligence.
Mainstream interest in AI and machine learning is at an all-time high, and the industry is responding by producing thousands of AI and ML studies for conferences and journals. The AI/ML community is also particularly active in posting non-peer-reviewed preprints via online platforms such as ArXiv. With this plethora of jobs, what is it that makes you rise to the top and get noticed?
The answer, at least in part, is: According to a survey conducted by two of her highly influential users: X (formerly Twitter) selected highlights, according to a new preprint from researchers at the University of California, Santa Barbara.
The UCSB paper analyzed more than 8,000 AI and ML papers, considering both social media mentions and citation counts. The researchers examined tweets from December 2018 to October 2023 and found that the median number of citations for AI/ML papers shared by two specific influencers was lower than the number of citations for the control group. It was concluded that it was 2 to 3 times as large.
This is very important because academic citations do not simply signify recognition in a person’s field. It also influences decisions such as research funding and tenure at academic institutions. And that’s a change from the status quo. As recently as 2018, a study of conference papers showed that a paper’s review score (meaning acceptance to a top conference) is a key indicator of future citations.
Well, this new study concludes, “The correlation between influencer tweets and citation counts, rather than review scores, points to a shift in how communities find and read papers.”
Two influencers who are having a huge impact on AI
The researchers selected two influencers as case studies. Both consistently share her AI/ML papers and have gained many followers on X (formerly Twitter). @_akhaliq and @arankomatsuzaki. “These influencers, similar to journalists in civil society, emerge as key figures navigating a flood of information, highlighting and setting context for works that matter to their communities,” the authors write. writing.
Given the deluge of research, that curation is a valuable, and of course free, service from these influencers. However, “over-reliance on a select group of curators may unintentionally distort the research landscape by emphasizing certain topics or perspectives over others,” the researchers wrote. . The paper adds that inadvertent bias in sharing research results from particular labs or researchers can perpetuate a lack of geographic, gender, and organizational diversity. .
Raising awareness is the first step to breaking out of this social media echo chamber, says the lead author. Iain Xie Weissburg, first-year master’s student in UCSB’s Electrical and Computer Engineering program.
“We wanted to help the community recognize this and be vigilant to ensure that research is at an even level,” he says. spectrum. “As it stands, we all tend to get our information from a select few, conclude that these are hot topics, and often select studies based on that hype.”
The key, Weisberg cautions, is not to shame these influencers or others or place undue responsibility on them. “This means that publishing and conferencing systems will need to adapt to a significant increase in the amount of AI/ML research, a trend that is expected to continue for the foreseeable future, especially with the influx of generative AI into the public realm. ,” says Weisberg.
This analysis highlights not only the growing influence of social media in AI/ML research, but also the importance of an evolving ecosystem for bringing diversity of thought to today’s digital academic environment.
The amount of AI papers is overwhelming.
Researchers say their methodology for selecting just two influencers is ‘far from perfect’ Derip Rao, independent researcher at the University of Pennsylvania and the University of California, Santa Cruz. Additionally, “These two guys tend to tweet papers from big labs and celebrities, so it’s not clear who is influencing whom.”
Still, he agrees with the overall conclusion that a small, influential group has outsized influence and is “problematic for science.” Citation counts are intended to be determined by experts who have a deep understanding of the work they’re citing and its context, he noted, adding that even expert influencers’ external social media activity of their day-to-day work, adding that it would be “unrealistic” to expect them to bring this kind of rigorous review to their day-to-day work.
This is especially true given the recent surge in AI/ML research, as one of two featured influencers recently acknowledged. “I hate how arXiv randomly releases 500 papers almost every day.” Alan KomatsuzakiI wrote to X.
“The sheer volume of papers published every day makes it impractical for individuals to sift through arXiv feeds,” said Komatsuzaki, chief technology officer at Teraflop.ai. spectrum In an interview. “The community relies on curators like @_akhaliq and me to highlight noteworthy papers every day…. [I’m] We are careful not to promote research with unconvincing, weak, or questionable results. ”
Another influencer mentioned, Arsen Khalik, or “AK,” is a machine learning engineer at Hugging Face. “I think there’s been some shift in the community towards finding new research, discovering it, and citing it through Twitter and other social media rather than conference and peer reviewer scores. But each has its own position within the community,” Khaliq said in an interview.
Will the AI social media bubble burst?
The paper’s core conclusion is that researchers, conference organizers, and academic institutions are changing norms as preprint platforms and social media accounts change the landscape of research sharing, especially in the AI/ML field. We need to recognize that, says Weisberg. The authors also argue that conference and peer review processes may need to evolve to effectively disseminate high-quality research.
Weisberg says that engineers working in this field should sometimes resist the urge to jump on the hype, lest they ignore other important areas of research. Influencers can also note that “paper sharing is not a zero-sum game.” You have influence, but you don’t always have to share the biggest companies or the most famous researchers. It’s important to have a diverse community with different ideas and backgrounds. ”
In future research, Weisberg hopes to explore the potentially huge impact of social media in other scientific fields. He also has the opportunity to investigate the underlying mechanisms of social media (such as how algorithms display content to users) as they relate to academic recognition.
Rao is looking forward to new ideas for publishing and disseminating high-quality papers. “When you have overproduction, it’s natural to turn to curators. The community’s dependence on such people is a cry for help,” he says. “We need a better way to combat this information overload, and the answer is preferably not influencers.”
