Recent research has found that sentences automatically generated by AI writing assistants or the "smart replies" they provide not only enhance people's creativity in expression but also influence their thoughts. Maurice Jakesch, a doctoral student in information science, asked over 1,500 participants to write a paragraph in response to the question, "Is social media beneficial to society?" Individuals who used an AI writing assistant with a bias towards social media were twice as likely to agree with the AI's suggestions compared to those who did not use AI, potentially indicating a shared viewpoint.
Researchers suggest that this study demonstrates how biases, whether intentional or unintentional, present in AI writing tools can impact culture and politics. "We are eager to implement these AI models across various industries, but we need a better understanding of their implications," says Mor Naaman, co-author and researcher at the Jacobs Technion-Cornell Institute. "In addition to improving efficiency and creativity, they may have other effects on individuals and our society, such as transforming language and perspectives."
While previous studies have explored how large language models like ChatGPT can create persuasive advertisements and political content, this is the first study to suggest that the process of writing with AI tools can influence a person's viewpoint. Jakesch presented this research at the CHI International Conference on Human-Computer Interaction in April 2023, stating that "collaborating with opinionated language models in writing can impact users' perspectives," earning the paper an honorary award.
To understand how people interact with AI writing assistants, Jakesch manipulated a large language model to have either positive or negative views on social media. Participants were asked to write a paragraph on a platform resembling a social media website, either on their own or in collaboration with an opinionated AI writing assistant. The platform collected data on participants' acceptance of AI suggestions and the time taken to write the paragraph.
According to judges' evaluations, participants who used the AI assistant supporting social media wrote more sentences advocating its benefits compared to those without an AI assistant, and vice versa. These participants also tended to express similar viewpoints to the AI writing assistant they used in subsequent surveys.
The researchers examined the likelihood of individuals accepting AI suggestions solely for the purpose of completing the task faster. However, even participants who spent several minutes writing their paragraphs presented significantly influenced viewpoints. The survey revealed that most participants did not even notice the biases in the AI or realize they were being influenced.
"I didn't feel persuaded during the collaboration," says Naaman. "It felt like I was doing something very natural and organic—I was expressing my own thoughts with some assistance."
When the experiment was repeated with different topics, the research team once again observed the influence of AI writing assistants on participants. Currently, the team is investigating how this experience alters people's viewpoints and how long-lasting these effects are.
Just as social media has changed political landscapes by facilitating the spread of misinformation and creating echo chambers, biased AI writing tools can also lead to similar shifts in perspectives, depending on the tool used by the user. For instance, some organizations have announced plans to develop alternatives to ChatGPT that aim to express more conservative viewpoints.
The researchers argue that these technologies warrant further public discussion, including how they can be abused and how they should be monitored and managed. "As these technologies become more powerful and deeply embedded in social structures, we should carefully consider how to manage the embedded values, priorities, and perspectives," says Jakesch.
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.