Main Body
The Attention Economy & Algorithmic Search
“As users engage with technologies such as search engines, they dynamically co-construct content and the technology itself.” – Safiya Umoja Noble
Overview
In the prior readings, you explored the theoretical construct of a mediatized communication ecology and the relationship to the human construction of reality.
In this chapter, we focus on the digital systems within our mediatized ecology and the algorithms that determine what you see as you engage with them. If you have ever studied information literacy in a prior course, you were likely focused on the content of information to determine its objective validity and reliability. Instead, in this area of study, we are interested in knowing how that information appeared in front of you as an output of your engagement with a system.
Why does this matter? It is because popular search engines like Google and AI are where many users go to confirm “the real truth” instead of relying on other media institutions or experts. To many, a Google or AI search is an operationalized method of “Doing your own research.”
The problem with reliance on search engines and AI is that there is a presumption that the top search results, most popular selections, or AI output affirm the validity of a given answer. From the user’s perspective, digital systems are presumed to be objective arbiters of fact checking (Sundin, O., & Carlsson, H., 2016). However, this is not the case. The results of search queries reflect a combination of factors including what other people have selected in similar search queries.
“Knowledge is not simply conveyed to users, but co-produced by the search engine’s ranking systems and profiling systems, none of which are open to the rules of transparency, relevance and privacy in a manner known from library scholarship in the public domain (van Dijck, 2010, p. 575).”
With AI systems trained on Internet data, there is a risk that the output of AI published on the Internet will be circulated back into its own source content, which results in a recursive spiral of “junk data.”
“We find that indiscriminate use of [AI] model-generated content in training causes irreversible defects in the resulting models, in which tails of the original content distribution disappear. We refer to this effect as ‘model collapse’ and show that it can occur in LLMs…. (Shumailov, I., Shumaylov, Z., Zhao, Y. et al., 2024)
When you trace the flow of information in algorithmically optimized systems, you find that the “truth” in digital network systems isn’t about the validity of information, it is about what the majority of people believe is true. As more people accept top search results and AI output as objective proof of a given body of knowledge, it creates a feedback loop that elevates popular content to a higher state of relevance regardless of its quality. In fact, search results are designed to be subjective and are, to a degree, determined by who you are, your interes