I. Introduction
Figure 3 - News should be outlined: placing every article in a living tree of issues
- A war of warrants: Debate victories can be vectorized as weights, so a crowdsourced decision-making AI is stress-tested against competing perspectives. The central claim is that news should be outlined, with each claim placed in a living tree of issues, sub-issues, citations and impacts.
- Inspectable collective thought: Language models distill collective thought into a weighted vector space, and because the system exposes which sentences and citations informed an output, users can check whether its reasoning matches their interests. Research Agents recommend articles, monitor topic keywords, and extract cited references from uploaded papers.
- Documents, not the debate: Google ranks pages but never shows where an article fits in the larger topic. A topic-level mind map, like a debate outline's tree headings, gives relevant highlights without an ideological bubble.
Debate should be a war of warrants where victories are vectorized as weights, so that a crowd-sourced decision-making AI can be publicly stress-tested against competing perspectives. Critical times call for a crowdsourced argument-reasoning dataset that lets AI models recommend research quotes, evolves chain-of-thought reasoning, unlocks faster reading of long articles, monitors developments through topic-modeled knowledge graphs, and provides a public answer service. The central claim is simple: news should be outlined. Instead of reading the internet as disconnected articles, users are able to place claims into a living tree of issues, sub-issues, citations, and impacts. In this model, debate becomes not only a pedagogical activity but also a synergistic structure for social sensemaking.
Language models can distill collective thought into a vector space in which each point carries a weighted contribution to decision-making. Because the system exposes which sentences and citations informed an output, users can inspect whether its reasoning aligns with stated interests. Research Agents recommend articles for human researchers working alongside AI to develop a summarized topic outline as a public service, while monitoring related articles via web searches for keywords associated with that Topic Model. When a user uploads a research paper, the system extracts the full text of cited references, generates topic-model and keyword summaries, monitors the associated literature, and stores highlighted passages. Users could maintain Personal Knowledge Bases that record influential sources, which a future assistant might use to model research interests.
Debate has traditionally trained users to compare claims, test warrants, and evaluate evidence under adversarial conditions. Recent work in AI debate and responsible AI in education shows growing interest in systems that assist argument generation, evidence review, and ethical reasoning while preserving human oversight (Wambsganss et al., 2024). This paper extends that trajectory by proposing a platform-centered model in which debate workflows become infrastructure for tracking current events, organizing public research, and aligning AI outputs with transparent evidentiary support. Google returns a ranked list of pages, but not a map showing where each article fits within the larger topic structure — it shows documents, not the debate. A topic-level view allows users to see where each article fits into the larger mind map of a topic, similar to a debate outline's tree headings, providing relevant highlights without a narrow ideological bubble mindset.
