THINK FIRST·CODE LATER

← All labs

Read an embedding map produced by an AI pipeline

Problem

A city chatbot logs citizens' questions. The vendor's analytics dashboard embeds all 50,000 questions with a language model and shows a UMAP map: points coloured by an automatically generated "topic" label (an LLM named each cluster), with a headline "Transport complaints form the largest and most isolated group".

  1. List what the map can and cannot support as evidence.
  2. Describe how you would verify the topic labels and the headline.
  3. Propose a set of views (at least three) that give officials a trustworthy picture of what citizens ask about, and describe how AI could still help.

Work it out on paper, in a document or here, then compare with the model answer. Your answer stays in your browser — it is never sent to or stored on the server.