View on GitHub

CS357

Foundations of Artificial Intelligence

CS357: Foundations of Artificial Intelligence - Team Charter and Norms

Purpose

To write your team’s constitution: rotating roles, psychological-safety ground rules, and a conflict repair process you agreed on before you needed it.

About This Tutorial

Today your new semester team writes its constitution. This tutorial is modeled on the Team Contract Exercise developed by the SWEET Center at Worcester Polytechnic Institute (WPI SWEET Center Team Contract Exercise), which has teams agree on norms, decision rules, and repair procedures before the first conflict rather than during it. We move from getting to know each other honestly → understanding how teams develop and what makes them safe → writing an enforceable charter you will all sign.

You drafted a small pre-draft charter in the Overview assignment. Bring it: you will expand it into the real thing. The signed charter is due at the charter milestone on the course schedule, and your team will formally revisit and revise it at the midpoint charter revisit on the schedule.

Purpose, Task, and Criteria

Following the TILT (Transparency in Learning and Teaching) framework (tilthighered.com):

   
Purpose To convert your new team from a set of strangers into a working community with explicit, agreed, enforceable norms, the collaboration foundation for the entire Project Thread (Goal 13), and a live answer to the Open Question How should we live together? (Goal 15).
Task Complete the three models below, then draft, discuss, and sign your Team Charter using the template in the Exercises.
Criteria A charter is done when every section passes the enforceability test (a stranger could tell whether the team is following it), every member has signed it, and it names when it will be revisited.

Key Concepts

Term Plain-English Definition Where You’ll Meet It
Team Charter A short, signed document recording the team’s norms, decision rules, conflict protocol, accountability procedures, and revisit clause. A contract with yourselves. The template you fill in during the Exercises
Norm A specific, observable behavior the team agrees to expect from everyone. Enforceable norms describe behavior (“respond within 24 hours on weekdays”); unenforceable ones describe attitudes (“be respectful”). “Agenda posted before every meeting” is a norm; “communicate well” is a wish
Tuckman Stages The developmental sequence most teams pass through: forming, storming, norming, performing, adjourning (Tuckman, 1965). Storming is a stage, not a malfunction. Model 2’s scenario trace
Psychological Safety A shared belief that the team is safe for interpersonal risk-taking: questions, mistakes, and disagreement do not get punished (Edmondson, 1999). The norm that makes every other norm workable
Decision Rule The agreed procedure for making a choice when the team does not naturally agree: consensus, majority vote, role-holder decides, or escalate. “Technical disputes: 10-minute timebox, then the sprint Coordinator decides and logs it”
Repair Move A named, rehearsed conversational move for de-escalating conflict and fixing a rupture: name it, own your part, restate the other side, propose a next step. Model 3’s conflict protocol
Revisit Clause The charter section stating when and how the charter itself gets re-examined and amended. “Full revisit at the midpoint charter revisit on the schedule; any member may call an early revisit by posting AGENDA: charter”

Part I: Who Is on This Team?

The Get-to-Know Protocol

Why this matters: Most team conflicts are not caused by malice; they are caused by unshared assumptions. The Team Formation Survey already asked each of you four questions privately. Now you answer them to each other. Structured self-disclosure like this is the fastest route through Tuckman’s forming stage, and it is the first deposit into psychological safety: each answer is a small interpersonal risk that the team practices receiving well (Edmondson, 1999).

The protocol. Go around the team four times, one round per prompt. One person speaks at a time; listeners may ask one clarifying question, but may not debate or “fix” anything. The Recorder keeps a grid like this (you will use it in Model 3):

Member Challenges Energy Sources Pet Peeves What Matters Most
A “Asking for help late” “Whiteboard brainstorms” “Silent channels for days” “Nobody surprised at the deadline”
B … … … …

Round prompts, in order:

  1. Challenges: What is hard for you in team settings?
  2. Energy sources: What kind of teamwork leaves you energized rather than drained?
  3. Pet peeves: What specific teammate behavior reliably frustrates you?
  4. What matters most: Complete the sentence: “For me, a team is working well when __.”

Questions to Work Through

  1. Look at your team’s completed grid. Find one pair of answers that could collide: for example, one member’s energy source is another member’s pet peeve (long unstructured brainstorms, say). What norm could you write now that lets both people get what they need?

Hint: The collision is not a problem to eliminate but a constraint to design around. A norm like “meetings have 20 minutes of open brainstorm, then 20 minutes of silent structured work” serves both the brainstormer and the person drained by open-ended discussion.

  1. The protocol forbids listeners from debating or “fixing” what a speaker shares. What does that rule protect, and what would likely happen in round 1 (Challenges) without it?

Hint: Think about what it costs a person to admit “I struggle to ask for help”, and what it teaches them if the immediate response is advice, judgment, or someone else’s story. Would round 3 (pet peeves) still get candid answers?

  1. Which single cell of your team’s grid do you predict will matter most by the midpoint charter revisit? The Recorder should write the prediction down, with a one-sentence reason; you will check it at the charter revisit.

Hint: Look for the answer that interacts with deadlines, since the thread’s crunch points (the stakeholder brief, the synthesis, and the proposal milestones on the course schedule) are where working-style differences surface.

Part II: How Teams Actually Develop

Tuckman’s Stages and Psychological Safety

Why this matters: Tuckman (1965) reviewed dozens of studies of small groups and found the same developmental sequence appearing again and again: forming (polite orientation), storming (conflict over goals, roles, and standards), norming (shared expectations settle in), performing (the team self-corrects and produces), and, in later work, adjourning (the team ends). Edmondson (1999) supplies the variable that determines whether storming produces growth or wreckage: psychological safety, the shared belief that the team is safe for interpersonal risk-taking. Teams with it surface problems early and learn; teams without it hide problems until they explode.

Here is the same idea two ways: a stage table, and then a scenario trace through a storm.

Stage What It Looks Like The Trap The Charter’s Job
Forming Politeness, vagueness, easy agreement Mistaking politeness for alignment Force specifics now (this tutorial)
Storming First real disagreements; frustration Reading conflict as “the team is broken”; going silent A rehearsed conflict protocol with repair moves
Norming Rhythms, shorthand, trust Norms drift undocumented The midpoint charter revisit codifies what changed
Performing Self-correcting, low-drama output Coasting on unexamined habits Lighter process, but the decision log continues
Adjourning The team ends; work is handed off Ghosting the ending Contribution statements and final reflection

A storming scenario: trace it. Read this and follow the questions below.

It is week 6. The team’s Literature Review annotations are due Friday. Member P posted their annotations Monday. Member Q has posted nothing and has not answered the channel in three days. Member R, frustrated, writes Q’s annotations themselves Thursday night and posts to the channel: “done, since apparently nobody else was going to do it.” Q finally appears Friday morning: they had a family emergency, they had told the Manager privately on Tuesday, and they had two of three annotations drafted offline. Q reads R’s message, says nothing, and stops contributing in meetings.

Questions to Work Through

  1. List every point in the scenario where information existed but did not reach the people who needed it. For each, name a charter norm (specific and observable) that would have routed it correctly.

Hint: There are at least three: Q’s silence in the channel (a check-in/response-time norm), Q’s private message to the Manager that stayed private (a norm about where absence information goes: the channel, not a DM), and R’s unilateral takeover (a missed-deadline procedure that says what happens, and when, before anyone rewrites anyone’s work).

  1. R’s Thursday-night message is the moment psychological safety takes damage, for Q, and for everyone watching. Using Edmondson’s (1999) definition, explain what every other member of the team just learned about the cost of falling behind, and why that lesson makes the team’s next problem more likely to stay hidden.

Hint: Psychological safety is a shared belief built from observed evidence. The audience for R’s message is not just Q. If struggling publicly earns public sarcasm, what will members do the next time they are struggling: surface it early, or hide it and hope?

  1. Write the repair script: what should R say at the Friday meeting, and what should Q say, if both follow the four repair moves (name it, own your part, restate the other side, propose a next step)? The Recorder writes both scripts word for word; your team will adapt them into your charter’s conflict protocol.

Hint: R might begin: “I want to name that my message Thursday was sarcastic and public, and I’m sorry; I was anxious about the deadline (own). I didn’t know about your emergency (restate comes after listening)…” A repair script is short. The hard part is that someone has to go first; your charter can name whose job that is (hint: the current Reflector is a natural owner).

According to Tuckman (1965), a team having its first serious disagreement in week 6 of the semester is:

Answer

A normal, expected developmental stage that healthy teams pass through

Common Misconception: Psychological safety does not mean “niceness” or the absence of disagreement. Edmondson (1999) is explicit that psychologically safe teams disagree more openly, not less; safety is what makes candid disagreement survivable. A team that never argues is usually a team in which people have stopped saying what they think. The goal of your charter is not to prevent storms; it is to make storms informative.

Part III: Writing the Charter

Charter Construction

Why this matters: The WPI SWEET Center Team Contract Exercise, on which this tutorial is modeled, has teams write their contract in terms of behaviors a third party could verify, the same enforceability standard the Governance direction of the Responsible AI in Practice assignment will later apply to AI systems. Here is the anatomy of a charter, shown two ways: as a checklist of required sections, and as a worked before/after example.

Required sections (your charter must have all six):

  1. Norms: observable behaviors: communication channel and response-time expectation, meeting cadence and preparation, standup rhythm, agenda/notes discipline (see the Team Playbook).
  2. Decision rules: how technical and non-technical decisions get made when consensus does not appear: the timebox, the tiebreaker, and where the decision is logged.
  3. Conflict protocol with named repair moves: what any member does when frustrated with another (step 1 is never “post sarcasm to the channel”), the four repair moves your team commits to (name it, own your part, restate the other side, propose a next step), and who convenes the repair conversation.
  4. Accountability: the missed-deadline procedure (what a member does, and by when, if they will miss a commitment), the definition of “done,” and how work is rebalanced without takeovers.
  5. Role rotation: the POGIL/sprint role schedule, mirroring the Playbook cadence (weekly through the midpoint charter revisit, per-sprint afterward).
  6. Revisit clause: the scheduled full revisit at the midpoint charter revisit on the course schedule, plus the trigger any member can pull to call an early revisit.

Before and after, the enforceability test:

Weak (unenforceable) Strong (enforceable)
“We will communicate openly.” “Course communication happens in the team Discord channel; weekday messages get a reply (even just ‘seen, will respond tomorrow’) within 24 hours.”
“We will resolve conflicts respectfully.” “A member frustrated with a teammate raises it privately with that teammate within 48 hours using the four repair moves; if unresolved after one conversation, the current Reflector convenes a 15-minute team discussion with a written outcome.”
“Everyone will do their share.” “A member who will miss a commitment posts to the channel at least 48 hours before the deadline with what is done, what is not, and what help they need; the team rebalances in the next standup. Nobody rewrites another member’s section without that member’s reply or a 24-hour silence.”

Questions to Work Through

  1. Take your Overview pre-draft charter and run every sentence through the enforceability test: could a stranger, reading only your team’s channel and documents, tell whether the sentence was being followed? Count the sentences that pass. What fraction survived, and what does the most common failure look like?

Hint: The most common failure is an attitude word doing the work of a behavior: “respectful,” “open,” “fair,” “flexible.” Each one needs to be traded for a who-does-what-by-when. This is the same third-party test you will apply to governance clauses later in the course; practice it here first.

  1. Your Model 1 grid holds each member’s pet peeves and challenges. Pick the two entries with the highest stakes and write one charter norm each that pre-empts them. Why is a norm written from a real member’s answer stronger than one copied from a template?

Hint: A template norm protects a hypothetical team; a grid-derived norm protects this one. It is also self-enforcing in a way templates are not: everyone remembers why the norm exists and who it protects, which makes invoking it feel like care rather than rule-lawyering.

  1. Decision rules trade off speed, buy-in, and quality. For each of these three decisions, choose a rule (consensus / majority vote / role-holder decides / escalate to instructor) and defend the choice: (a) which stakeholder your team pursues for the Brief, (b) which retrieval library to use in a sprint, (c) whether to report a teammate’s sustained non-participation to the instructor.

Hint: (a) is high-stakes and identity-flavored; buy-in matters most, which argues for consensus with a deadline. (b) is reversible and technical; speed matters, which argues for a timebox plus role-holder decision, logged. (c) involves the charter itself failing, which is exactly what escalation paths are for, and why the charter should name the threshold in advance so reporting is a procedure, not a betrayal.

Which of the following charter clauses passes the enforceability test?

Answer

“The Recorder posts meeting notes with action items and owners to the shared folder within 24 hours of each meeting.”

Exercises

Draft the charter (30-40 minutes). Using your Overview pre-draft as the starting point and your Model 1 grid as raw material:

  1. Draft all six required sections. Every member is primary author of at least one section (name the author under each section heading), the standing Project Thread rule for all team documents.
  2. Exchange a one-minute read-aloud of the conflict protocol with a neighboring team and collect one suggestion each way.
  3. Run the full enforceability pass (Question 7) on the final text as a team.
  4. Sign it.

Deliverable statement. The finished charter must be signed by ALL team members (typed names with date), and it is a living document: every future progress report your team submits (sprint reports, milestone submissions, check-ins) carries all members’ signatures, re-affirming the charter each time. Submit the signed charter by the charter milestone on the course schedule as directed in the LMS. Your team will formally revisit the charter at the midpoint charter revisit on the schedule: re-read it, mark each clause held / drifted / never real, and file a signed revision; teams change, and a charter that cannot change with them stops being true.

Reflection Prompt

In your notebook, keyed to Open Question 2: How should we live together? Your charter is a small constitution: it decides how power (decisions), voice (norms), and mercy (repair) work in a five-person society for one semester. Which clause of your team’s charter would scale to a community you actually live in (a house, a club, a workplace) and which clause only works because the semester ends? What does the difference tell you about what living together well requires when there is no adjourning stage?

Where This Goes Next

Your team exists and has a constitution. Next on the Project Thread, you point it outward: identifying a real stakeholder outside computer science and writing the Stakeholder Brief, the problem your whole semester will orbit.

Further Reading