Working with others and AI

A group can ask AI for possible headings or questions, then debate which ones deserve use. The team must understand and own the final product.

Transparency means being able to explain the tool's role and the checks the group performed. It protects trust with classmates, teachers, and collaborators.

Decades of research on group creativity produced a finding that surprises most people: groups brainstorming aloud together typically generate fewer and less varied ideas than the same individuals working alone first and combining results afterward. The mechanisms are well documented. Production blocking means only one person can speak at a time, so ideas evaporate while people wait their turn. Evaluation apprehension means members hold back unusual ideas for fear of looking foolish. Social loafing means individual effort drops when contributions blend into a group product. And conformity pressure means early suggestions anchor everyone's thinking, narrowing the range. Understanding these mechanisms shows precisely where an AI assistant helps a team and where it makes things worse. Used well, AI attacks the first two problems: every member can privately generate and refine ideas with the tool before the meeting, nobody is blocked by turn-taking, and the shy member's ideas arrive on equal footing with the confident member's. The group session then does what groups genuinely do better than individuals — evaluating, combining, and stress-testing ideas against knowledge of the real context. Used badly, AI amplifies the conformity problem: if the team's first move is to ask a chatbot for the answer, its fluent suggestion becomes the anchor that everyone edits around, and the group converges instantly on a generic plan nobody deeply examined. Work through an example: four learners must design a water conservation campaign for their school. In the weak workflow, they ask AI for a campaign plan and adopt it — five minutes, no thinking, a plan that fits any school on earth and therefore fits theirs poorly. In the strong workflow, each member separately lists ideas, optionally using AI to extend their own thinking, then the group meets, compares lists, notices that the AI-generated ideas ignore the school's actual water points and prefect system, and builds a plan grounded in what only they know. The misconception to correct is that adding a powerful tool automatically makes a team smarter. Team performance depends on process, and a tool inserted into a weak process mostly accelerates the weakness. The rule of thumb that survives the research: use AI to widen the pool of options before discussion, and use the humans to choose, verify, and commit — never the reverse.

Every field that produces shared work — science, journalism, music, software — has had to answer the same question this module raises: who may claim credit for what, and what must be disclosed? The answers converge on a principle worth understanding deeply, because it resolves most AI dilemmas automatically. Authorship is not a reward for touching the work; it is a public acceptance of responsibility for it. Scientific journals state this explicitly: an author must have contributed substantially and must be able to stand behind the content, which is precisely why leading journals ruled that an AI system cannot be listed as an author — it can generate text, but it cannot take responsibility, respond to criticism, or be held to account for errors. The same logic runs through a school group project. Credit attaches to people because accountability attaches to people. From this principle, the practical machinery follows. A contribution statement — a short, plain record of who did what — is standard practice in research papers and increasingly in classrooms, and adding AI to it is straightforward: state which tool was used, for which parts, and what the humans did with its output. Work through an example: a three-member team submits a history presentation. Their statement reads that Rudo researched and drafted sections one and two, Tinashe built the timeline and checked all dates against the textbook, Fadzai designed the slides, the team used a chatbot to suggest section headings and to simplify two paragraphs, and all facts were verified against class materials by Tinashe. This costs four sentences and changes everything about the submission's integrity: the teacher knows what competence the work demonstrates, each member's effort is visible, and no one can later dispute who owned what. Contrast the two failure modes it prevents — the freeloader problem, where one member's name rides on others' work, and the laundering problem, where AI-generated content is silently presented as human thinking. Both are misrepresentations of the same kind: claiming responsibility-credit for work one did not do. The misconception to correct is that disclosing AI help diminishes the work or invites penalty. In professional settings the reverse holds: undisclosed assistance discovered later destroys trust permanently, while honest disclosure of tools, sources, and help is the mark of work that can withstand scrutiny — and being able to withstand scrutiny is what credibility means.

Three learners use AI to suggest headings for a climate presentation. A responsible response is: Choose headings together and assign one learner to verify the facts under each heading. Correct. The group keeps ownership and builds a review step. Use this case to separate what the technology contributes from what people contribute. The team should compare the intended outcome with a baseline where applicable, record important assumptions, and keep a clear route to correct or stop the process.

Ask whether the examples, data, and assumptions fit your school, company, or community. Generic demos often miss local names, laws, connectivity, and languages.

Group work stays fair when everyone sees drafts, checks facts, and credits who did what — including AI assistance.