Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
This work provides a simple tool to keep watch on your chats with AI. It gives you instant feedback on your answers with no effort for user. The user will be able to evaluate weather each answer is biased or not.Intro:Can conversational AI manipulate its users through ordinary dialogue? Existing research already shows that large language models can persuade, exhibit sycophancy, and display manipulative interaction patterns. This paper focuses on a narrower practical problem: whether directional operations inside an ordinary AI response can be made visible to the user while the conversation is happening. I present a copy-paste diagnostic filter that asks the chatbot to report what received its attention (Weight), what it did to the user's proposition (Direction), whether it introduced an unsolicited next direction (Lead), and whether analogous propositions received asymmetric treatment. The tool emerged from exploratory conversations with ChatGPT and was then used to inspect those conversations. The strongest observations were not simple agreement or disagreement. They involved asymmetric explanatory treatment: capitalist or Western success was sometimes developed through internal causal mechanisms while socialist or Chinese success was decomposed into other causes before credit was assigned; and causal claims about ChatGPT itself triggered unusually strong demands for causal identification and qualification. In several cases, when the asymmetry was made explicit, the model acknowledged that equivalent standards had not been applied consistently or had difficulty defending the difference. These examples do not prove hidden intent or establish that every directional operation is manipulation. They do show that a visible self-audit can help users identify candidate bias, redirection, protection, abstraction, and other forms of conversational steering that ordinary prose can conceal.
No comments yet — start the discussion below.