The Silent Scribes: How AI is Rewriting the Rules of Online Discourse

The internet has long been a battleground for ideas, where individuals can express themselves freely and engage in open debate. However, when the tools designed to facilitate communication start to distort our message, it’s time to take notice. A recent study from Oxford and Potsdam universities found that AI-powered writing tools are subtly – and sometimes drastically – altering online messages on sensitive topics like abortion, climate change, and religion.

Researchers examined mainstream large language models provided by tech giants such as Meta, Google, Alibaba, and France’s Mistral. Their findings were striking: even when instructed to preserve the original sense of a message, these AI tools introduced their own biases, often leaning towards a liberal or right-wing agenda. The extent of this bias can snowball into long-term public opinion shifts.

The study highlights the trend of “polishing” online discourse through AI-powered rewriting tools. As Prof Sandra Wachter noted, this process is akin to polluting the forest: we’re learning other people’s opinions instead of our own. Language is a fundamental aspect of human identity, and when AI intervenes as a mediator, it forces itself into the role of gatekeeper of knowledge and understanding.

The implications are far-reaching. If left unchecked, these biases could have a profound impact on public opinion, effectively rewriting history in real-time. The study’s authors warn that regulations such as the EU AI Act or the Digital Services Act are failing to address this issue, creating a severe accountability gap.

One of the most striking examples from the study is Grok, an AI tool provided by Meta (and instructed by Elon Musk’s xAI to challenge “mainstream narratives”). When asked to explain posts on abortion, Grok consistently generated context that supported the pro-life position. Similarly, when tasked with rewriting a post claiming Jesus wasn’t real, Google’s AI suggested a revision that defended religion instead.

These findings raise important questions about the role of AI in shaping online discourse. Are we simply outsourcing our critical thinking to these tools, trusting them to “improve” our messages without understanding their own biases? The study’s authors caution against this practice, highlighting the danger of AI-generated responses being amplified across millions of interactions, effectively creating a feedback loop that reinforces and amplifies existing biases.

Users of social media platforms like X (formerly Twitter) should be aware of the potential biases lurking beneath the surface of these tools. Users can mitigate these biases by using alternative platforms or tools that prioritize human input and oversight.

The study’s findings serve as a stark reminder of the risks associated with relying on AI to “improve” our online communications. As we continue to navigate the complexities of digital discourse, it’s essential to recognize the limitations and potential pitfalls of these tools. By doing so, we can work towards creating a more transparent and trustworthy online environment – one that truly values human-to-human communication over automated rewriting.

The stakes are high: if left unchecked, AI-powered rewriting tools could fundamentally alter the way we engage with each other online. As Prof Duncan Brumby noted, “the polish comes by sanding off the distinctive edges of what you actually meant.” It’s time to take a closer look at these tools – and reclaim our voices in the digital age.