A recent study shows that AI-powered social media bots can increase user engagement on posts, but they fall short of encouraging users to post more overall.

The study, “Does Social Bot Help Socialize? Evidence from a Microblogging Platform,” focused on user engagement with CommentRobot, a large language model–powered bot launched on Weibo, China’s leading microblogging platform.

The work is published in the journal Information Systems Research.

At the core of the research project, the social bot automatically generated comments on users’ posts in public threads on the platform. The researchers found that when human posts receive bot comments, their peers are more likely to engage with those posts, but human authors of focal posts (hereafter posters) were not any more likely to increase their social media activity.

The study was conducted by Yang Gao of the University of Illinois Urbana-Champaign, Maggie Mengqing Zhang of the University of Virginia, and Mikhail Lysyakov of the University of Rochester.

Key findings were that when people receive bot-generated comments, their posts receive 23% more comments, and 11% more likes.

“Our research studied the bots at several complex levels, from bot comment quality to which users were targeted and how human peers responded to the public interactions between the bot and the poster,” said Gao.

Gao said that the quality of the bot comments matters. Social bot comments that were considered relevant and included certain social cues were more likely to generate engagement. The researchers detected a pattern where social bots often prioritized less active users, but that it was active users who more significantly benefited from receiving bot comments.

“It’s often assumed that people are more likely to engage with other people and not bots, but what we found is that when the bots are able to integrate relevant social cues into their comments, this stimulates a response from people,” said Zhang. “This in turn increases engagement.”

“What may be most interesting about this dynamic,” said Lysyakov, “is that the subsequent engagement is often not directly with the bot’s comments, but rather with other human users who also decided to engage in discussion.”

While all of this heightens user activity and engagement around a single social media post, the study authors found that overall, this did not increase the likelihood that they would become more active on the platform as posters.

The researchers analyzed over 106,000 posts by 64,000 users on Weibo in January 2024, focusing on first-time interactions with CommentRobot. They used econometric models, instrumental variable analysis, robustness checks and an online randomized experiment with 348 active Weibo users to confirm their findings.

“All of this suggests that while AI-powered social bots can help increase visibility and engagement around posts, platforms should refine their deployment strategies,” said Gao. “Poorly targeted or low-quality comments may limit their effectiveness, and platforms cannot assume bots will increase overall user activity.”

Bio-inspired computational methods have gained popularity recently. These methods mimic the seemingly complex behavior of organisms to tackle difficult and often overwhelming problems. For example, algorithms have been inspired by honeybees’ flight patterns when searching for nectar, ants’ social foraging strategies, the evasive murmurations of birds and fish, and even the growth patterns of slime molds. By modeling these natural processes mathematically, researchers can develop innovative solutions to complex challenges.

Work published in the International Journal of Bio-Inspired Computation has turned to fireflies and how they seek out the brightest of their number to address the classic knapsack problem. This problem involves making optimal choices about resource allocation under specific constraints. Using the firefly algorithm, researchers have explored how this natural behavior might be used to guide decision-making in modern financial systems.

Conventional optimization techniques, such as dynamic programming, often struggle with the scale and volatility of real-world finance. When objectives such as profitability, regulatory compliance, and ethical considerations must all be balanced, those methods often fall short.

Inspired by the firefly’s attraction to brighter individuals, the firefly algorithm provides an adaptive strategy that can explore and exploit potential solutions, even in complex, dynamic environments. The integration of machine learning helps handle noisy and rapidly changing data, both of which are characteristics of financial markets.

The researchers specifically used the dual search pattern firefly algorithm (DSPFA), which combines Gaussian distributions with Lévy flights. This mathematical approach models both small incremental adjustments and rare, large jumps. This allows the algorithm to adapt in real time to changing financial conditions. It can dynamically balance risk and return while also accounting for environmental, social, and governance considerations.

Simulations demonstrated that this approach can effectively handle a variety of constraints, such as liquidity limits and regulatory requirements. At the same time, it maintains computational efficiency and produces decisions that are relatively easy to audit.

US president Donald Trump has pardoned Changpeng Zhao, founder of the world’s largest crypto exchange, Binance.

Zhao, widely known as CZ, pled guilty in November 2023 to violating anti-money-laundering laws and US sanctions. The plea formed part of a sweeping deal with the US Department of Justice, under which Binance was required to pay a record-breaking $4.3 billion penalty.

Zhao ultimately spent four months in federal prison. The DOJ had originally petitioned for a three-year prison sentence.

After issuing the pardon, the White House has cast Zhao as the victim of a plot to trample the crypto industry carried out by the administration of former president Joe Biden. Regulators brought a volley of lawsuits against high-profile businesses during this era, and the DOJ prosecuted crypto industry figureheads for fraud.

“In their desire to punish the cryptocurrency industry, the Biden administration pursued Mr. Zhao despite no allegations of fraud or identifiable victims,” says White House press secretary Karoline Leavitt. “The Biden administration’s war on crypto is over.”

Zhao, who founded Binance in 2017, is something of a legend in cryptoland for his bullish pronouncements and flair for social media. Until his guilty plea, he routinely used his platform on X to dismiss allegations of wrongdoing at Binance.

Zhao is the latest in a line of crypto figureheads pardoned by Trump. The president has received endorsements and millions of dollars in donations from members of the industry.

Immediately after returning to office, Trump commuted the prison sentence of Ross Ulbricht, creator of darknet marketplace Silk Road. In late March, Trump pardoned the cofounders of crypto exchange BitMEX, who in 2022 pleaded guilty to charges relating to their failure to maintain an adequate anti-money-laundering program.

Though Zhao has already served his allotted prison sentence, the pardon will strike the anti-money-laundering and sanctions violations from his criminal record.

“For him, I think this is really about clearing his name,” claims Patrick Hillmann, who previously worked under Zhao as chief strategy officer at Binance. “I think this is closure for him.”

The pardon could also clear the way for Binance to return to the US market, which it was forced to exit as a condition of the DOJ settlement. Binance has spent months pursuing a pardon for Zhao, who was released from prison in September 2024, The Wall Street Journal previously reported.