The Chatbot Optimization Game: Can We Trust AI Web Searches?

Civic Technology Community Group,

Hello. I would like to share a news article relevant to topics discussed in this group:

The Chatbot Optimization Game: Can We Trust AI Web Searches?
https://www.theguardian.com/technology/2024/nov/03/the-chatbot-optimisation-game-can-we-trust-ai-web-searches

The article broaches concerns about unfolding AI-enhanced search-related products and services.

Do any other caveats, concerns, or points come to mind?


Previously, we discussed the importance of preserving and protecting local journalism in villages, towns, and cities. One strategy, in these regards, involves equipping journalists with new tools and technologies.

As considered, interested customers (journalists, scholars, or scientists) should each be able to select one or more villages, towns, or cities (each region could be a distinct product) to subsequently receive personalized feeds of items resulting from the processing of the continuously-arriving content from each. These capabilities would, in part, be useful for solving the problem of expanding local news deserts [1].

Artificial-intelligence systems could process bulk content to send customers discoveries of specific interest to them in the form of extensible syndication items (e.g., RSS or Atom). Each syndication item could include one or more personalized explanations of why a customer would want to click upon its hyperlink to more fully explore content.

On a technical note, these explanations could be natural-language text or could contain hypertext or other markup languages, potentially referencing other items or other clusters of items of previous interest to a customer. These explanations could be generated on servers, on client devices, or by combinations of both. Abstract boilerplates or templates, for instance, could be generated on servers for subsequent processing, expansion, and completion by client-side artificial-intelligence systems, e.g., customers' virtual assistants.

As a result of AI-enhanced push-based information-retrieval and related technologies, over the course of time, customers would decreasingly have to attend, read through, or watch recordings of bulk public meetings, and would decreasingly have to read through bulk reports, documents, spreadsheets, presentations, and dataset descriptions.

Ideally, each village, town, city, county, state, and federal government would be of interest to multiple customers each making use of multiple competing services. By making use of multiple competing services, the amount of trust required from each customer for each service could be reduced. Streams or feeds from multiple competing services could be measured, analyzed, evaluated, and compared to one another (and, perhaps, to results from other search products and services) by interested customers, journalists, scholars, and scientists.


Best regards,
Adam Sobieski

[1] https://www.medill.northwestern.edu/news/2024/medill-report-shows-local-news-deserts-expanding.html