Ever since I read about the “online bubble” of google I decided to step out of that. I am in the internet to enrich myself, I need to hear opinions that are different from mine to grow up.

That was a long time ago, among other things what I did was browse without login in, remove cookies after closing the browser and use several search engines not just one.

Now I selfhost a searxng instance.

Researchers found that people often use search engines in ways that unintentionally reinforce their existing beliefs. The study, published in the Proceedings of the National Academy of Sciences, shows that even unbiased search engines can lead users into digital echo chambers—simply because of how people phrase their search queries.

"When people look up information online—whether on Google, ChatGPT or new AI-powered search engines—they often pick search terms that reflect what they already believe (sometimes without even realizing it),” said lead author Eugina Leung, an assistant professor at Tulane’s A. B. Freeman School of Business. “Because today’s search algorithms are designed to give you ‘the most relevant’ answers for whatever term you type, those answers can then reinforce what you thought in the first place. This makes it harder for people to discover broader perspectives.”

For example, people who believe caffeine is healthy might search “benefits of caffeine,” while skeptics might type “caffeine health risks.” Those subtle differences steered them toward drastically different search results, ultimately reinforcing their original beliefs.

The effect persisted even when participants had no intention of confirming a bias. In a few studies, fewer than 10% admitted to deliberately crafting their search to validate what they already thought, yet their search behavior still aligned closely with their beliefs.

The researchers tested several ways to encourage users to broaden their views. Simply prompting users to consider alternative perspectives or perform more searches had little effect. However, one approach worked consistently: changing the algorithm.

When search tools were programmed to return a broader range of results—regardless of how narrow the query was—people were more likely to reconsider their beliefs. In one experiment, participants who saw a balanced set of articles about caffeine health effects walked away with more moderate views and were more open to changing their behavior.

Users rated the broader results equally useful and relevant as the narrowly tailored ones. The findings suggest that search platforms could be crucial in combating polarization—if designed to do so. The researchers even found that most people were interested in using a “Search Broadly” feature—a button (conceptualized as doing the opposite of Google’s current “I’m feeling Lucky” button) that would intentionally deliver diverse perspectives on a topic.

Yeah, the way search engines have implemented LLMs have made it much worse.

You can go on google or duckduckgo and search for one thing, click the top link and click back after a few seconds. And the order of results has changed. So the second link you thought might have been correct(and was), is now 5-6 or lower. And the new second link is even worse than the first and usually generative content catered to the first link you clicked(wrongly).

I believe it, but I’m still debating whether something like Kagi is worth paying for. On principle, I strongly feel like it is, but in practice I’m still evaluating. So far, I’ve played with it a few times and I haven’t observed any notable improvements, but I’m trying to keep an open mind. First impression is that it’s definitely a little quicker and cleaner to get at the information I’m looking for. And taking a step back, I have to say it’s impressive that they can replicate a behemoth like Google’s accuracy already. On the other hand, I’ve felt like Google has gotten so crappy at search recently that maybe I’m simply not going to be satisfied with anyone simply “meeting” them and maybe what I want simply isn’t possible, in which case I’m just paying for disappointment.

Veritasium (youtube) did a piece on this, can’t remember which video it was, might’ve been the one on his thesis of how to teach STEM topics in videos

I wish they’d offer a way to mitigate bias in the summary article since we can’t see if in the actual study requires access to see if the researchers did

I haven’t used Google products or services in a decade!

…but I know everyone I know has…

Don’t tell me caffeine has drawbacks. Don’t wanna hear it!