Twelve Exercises from the Manual
12 exercises a reader can run on themselves, on a community they belong to, or on a project they are considering. Each can be done in 5-30 minutes alone, or in 30-60 minutes with a partner or small group. Drawn from the take-aways at the end of Chapters 10, 11, 12, and 13. Suitable for personal use, for the Wednesday salon, or as a workbook companion to the book.
Pick one exercise. Do it once. Give yourself permission to do it badly the first time.
The book’s central claim is that civility online is what each tribe protects, not what each tribe is. Both rooms in the corpus held 8-out-of-10 civility on housekeeping questions for almost three years and melted down the moment the prompt named an outgroup. The same user shifted 1.46 civility points between rooms. These exercises are how you find out which idol you protect when no one is watching.
If you can only do one, do A4. If you run a community, do B1. If you are about to launch, do C1. Otherwise start at A1. A pen and 30 minutes is enough.
For self-directed use on your own posts and your own online behavior.
Pick a post you are about to publish on a contested topic. Do not post it. Save it as a draft and wait 24 hours.
Re-read it cold the next day. Score it on the BTL test from Chapter 11. Three yes-or-no questions.
Two yeses and one no is a pass. One yes and two nos is the dehumanization lane. Rewrite or kill it.
The exercise takes five minutes the next day. The hard part is the 24-hour wait. Most posts that fail were written hot. The 24 hours is the real mechanic. If you cannot wait, that is also data.
Find a post you wrote 3 to 5 years ago about a contested topic. Search your old timeline, your old comments, your old emails. Read it cold, as if a stranger you do not know wrote it.
Ask one question. Would I find this writer trustworthy on other topics, or would I mark them as a member of a tribe and discount the rest of what they say?
The point is not shame. The 2017 author of the agnostic site_text essays wrote 36 of them and would now revise about half. The point is to learn what your tribe-protection signature looks like to the kind of reader you were trying to reach.
Pass: you can name, in one sentence, what tribal signal a stranger would read off your old post. Fail: you cannot, because you still think you were just stating the obvious.
Most readers fail this exercise on the first post they pull and pass on the second.
Pick two online communities you are in. They should be meaningfully distinct rooms. A professional Slack and a personal group chat. A subreddit and a hobby forum. A LinkedIn feed and a Discord. Two rooms where the default voice is different.
Pull one recent post you wrote in each. Score each on a 0-to-10 scale on four dimensions. Civility. Warmth. Argumentativeness. Self-disclosure.
The 108 cross-site members in the book averaged a 1.46-point civility drop, a 2.04-point warmth drop, and a 1.89-point argumentativeness rise between rooms. 85% shifted meaningfully without noticing.
Add the absolute differences across the four dimensions. That is your gap.
Pass: total under 4. The rooms are not writing your voice for you. Middle: 4 to 8. The rooms are shaping you and you should know which is doing what. Fail: over 8. One room is writing your voice and the other one would not recognize you.
List the 5 topics that, when raised online, cause your civility to drop the fastest. Be specific. Not “politics.” Specific topics. “Anything about Israel and Gaza.” “Anything about a former friend who became famous.” “Whether my city is safe at night.”
The book’s data shows each tribe’s purity reflex is outgroup-triggered, not constant. The agnostic room scored 9 on housekeeping prompts and 2.5 on a Trump-incitement prompt six days after January 6. Same room. Same week. Same people. Different trigger. The IDW room ran the same pattern with the inputs flipped.
Identifying YOUR triggers is the prerequisite for managing them. Until you have the list you will keep believing you are calm online and being surprised by your own posts.
Pass: 5 specific triggers, named without flinching. Fail: you stop at 2 because the next 3 are too uncomfortable to write down. The fail is the diagnosis. Write the next 3 anyway.
The list goes in your own pocket. It does not go online.
For each topic on your A4 list, write one sentence. What specifically am I afraid of losing if I let myself be persuaded by the other side on this question?
Status with my friends. My sense of who I am. The argument I have been making for ten years. The job I got hired for. The community I belong to.
Most readers find this harder than expected. The honest answers are usually about belonging, not about the topic. That is the finding the book points at and does not solve.
Pass: 5 sentences, one per topic, all of them honest. The exercise is private. You will know if you cheated.
For people running or moderating an online community, a salon, a discussion group, a Discord server, a Substack with comments.
Take 30 random posts from your community in the last 30 days. Read them in one sitting.
If a stranger had to name the cluster behind those posts, what verb would they use? Looking-for, or pushing-back? Seeking, or contrarian?
The agnostic room sorted into 8 clusters with names like Bruised Seekers and Tentative Truth-Seekers. The IDW room sorted into 6 clusters like Heterodox Traditionalist Skeptics and Cultural Grievance Veterans. Neither verb is wrong. The verb determines who comes back the second day.
Compare the cluster name to your stated mission and your written welcome script. If they do not match, the welcome script is the lie.
Pass: the verb in your welcome script is the same verb that describes what new members find. Fail: the welcome promises “thoughtful conversation” and the cluster name is something sharper.
If you fail, fix the welcome script first, then fix the room. Doing those two in the wrong order is how the book’s rooms ended up with mismatched populations.
Look at the last 50 posts in your community. Identify the writers. Count distinct users who account for the top 80% of posts.
In the book, about 6.9% of agnostic active users wrote 92% of responses on contested-thread cross-posts. About 13.6% on the IDW side did the same. Most rooms run 5 to 15% loud and 85 to 95% quiet.
The exercise is one number. The count of writers who carry 80% of your room.
The community tone is the loud minority’s tone, not the median member’s. If your loud minority is 5%, the room sounds like 5% of your members. The other 95% read more than they post and adjust their behavior to whatever the 5% set as normal.
Pass: you can name the loud writers by handle and you have read their last 10 posts each. Fail: you assumed the room sounded like the average member. Plan accordingly.
If your community has internal conflict, look for the cluster-pair pattern.
On agnostic, one specific cluster pair carried 702 reports between them, 20.6 times the off-diagonal mean. Two groups in reciprocal combat while the rest of the room read past the fight. The analogous IDW pair carried 49 reports, 5.4 times the mean. Both shapes were two-sided.
For your room, list the top 5 reporters and the top 5 reportees in the last 90 days. How much overlap is there?
High overlap means concentrated conflict. The moderation lever is targeted. A handful of named users in reciprocal combat. Talk to them individually.
Low overlap with scattered reports means diffuse conflict. The lever is structural. Change the prompts, the welcome, the format.
Pass: you can name the top conflict edge by handle pair. Fail: you have been running general moderation policy when the actual conflict is between four people.
List the three topics where members reliably break the civility rules they otherwise agree to. Not the topics with the most posts. The topics with the most rule-breaking per post.
Those are the idols the room protects.
Whether those idols deserve protection is a separate question, and not the one this exercise is for. The first job is naming them honestly. Civility is what each tribe protects. Every tribe is protecting something specific. Your room is not the exception.
Pass: three topics, named in concrete language. “Race in city policy,” not “social issues.” “Founders we used to admire,” not “personal stuff.”
Fail: you list one and stop because you can already feel the room’s reaction to seeing the other two on a list. The reaction is the proof the topics belong on the list.
For founders, builders, philanthropists, organizers considering whether to build a new room.
The exact test from Chapter 11. Run it on a project you are considering building, funding, or joining.
Three pass-or-fails.
Pass: cleanly first-half on all three. The project is in the rehumanization lane. Build it.
Fail: cleanly second-half on any one. The budget would be better spent on coffee for somebody else’s room.
Most projects land in the middle and need careful pulling toward the first half to be worth the time. The exercise tells you which way to pull.
One question. Will this project work better at scale or by repetition?
Most rehumanization rooms scale by repetition. Weekly. Same people. Slow on purpose. The Wednesday salon at 110 Union Street is a repetition project. Three to seven people, same room, every week, for years.
Most dehumanization platforms scale by virality. One-shot. Infinite people. Fast. A viral Substack post is a scale project.
Both are fine. You cannot do both with the same product. Pick one before you launch. If you pick repetition, plan for it to look small for the first 6 months. If you pick scale, accept that the room will not hold a hard question without melting down.
Pass: one clean answer. Fail: “both.”
One question. Are you willing to pour the coffee yourself, every Wednesday, for 6 months, while wondering whether the project is working?
The first few Wednesdays at 110 Union Street had three people in a 16,000 square foot room, two of them on payroll. Eighteen months later it is the room you read about in Chapter 12. The difference was somebody pouring the coffee on the Wednesdays in between.
If you are not willing to pour the coffee yourself for 6 months, the project is not yet what you think it is. Hire a fundraiser, an operator, a producer, but do not hire someone to pour the coffee. The coffee is the project.
Pass: yes. Fail: anything other than yes.
The point of these exercises is not to score yourself well. The point is to know what you are protecting.
The book’s central claim is that civility online is what each tribe protects, not what each tribe is. These 12 exercises run that test at the individual, community, and project level. Most readers will fail two or three on the first pass. That is the right outcome. The exercises that hurt to do are the ones that taught you something.
Stop by the salon any Wednesday and do them with us. 110 Union Street, downtown Seattle, 4 to 7 pm, free, coffee on me.
Art Love Salon, 110 Union Street, downtown Seattle, across from the Seattle Art Museum. Wednesday 4 to 7 pm. Free. Coffee on me.
Visit the Art Love Salon