Codes of Conduct That Work
Most conference codes of conduct are copied, and copying is the correct thing to do — writing one from scratch produces something worse. For a commercial example of related workforce measurement, see employee attendance tracking software from Monitask.
What separates a working code from a decorative one is not the text. It is whether anybody can act on it, and that part cannot be copied. For broader conference and speaking context, EuroPython is another useful reference.
Our 2016 code was based on the PyCon UK 2015 code and released under a free licence; it has since been reused by other conferences. It is preserved in the archive.
The text is the easy part
Copy from an established conference, adapt the specifics, credit the source. PyCon UK, the Django community and several foundations publish codes explicitly intended for reuse.
What must be adapted: contact details, the reporting mechanism, the jurisdiction, and anything referencing the venue.
What should not be softened: the list of unacceptable behaviour. Vagueness in that section is what makes a code unenforceable, and it is the section people are most tempted to make diplomatic.
And release yours under a free licence. Ours was, and other conferences reused it — which is how the ecosystem improves without every organiser writing prose about harassment.
What cannot be copied
A named person who takes reports. Not "the organisers" — a name and a way to reach them during the event, printed on the badge.
A second person, of a different gender where possible, because a reporter may not be comfortable with the first.
A decision procedure agreed in advance. Who decides, how fast, and what the range of responses is. Deciding this during an incident produces the worst version of every choice.
And a willingness to act. The one thing that makes the difference and the one that cannot be written down beforehand.
The decisions to make before the event
Written, agreed by the organising team, not published.
Who can eject someone. One or two named people with the authority to act immediately, without convening anybody.
What the range of responses is. A word, a warning, removal from a session, removal from the event, a report to an employer. Knowing the range in advance stops the response defaulting to the mildest available.
What happens if the person reported is a speaker, a sponsor or an organiser. These are the cases that break unprepared teams, and they are not rare.
And what is recorded. Something happened, when, what was decided. Not for publication — for the next team, and for the case where a pattern emerges across years.
What we got right and wrong
Right: a code adapted from a serious source, published before the call for proposals, with named contacts.
Wrong: we had not agreed the decision procedure in advance. The escalation path existed in our heads and not on paper, and the difference only matters when the situation is stressful — which is precisely when it arises.
And we did not write anything down afterwards, which is the same failure as everywhere else in organising: the knowledge stayed with the people who were there.
Sponsors and speakers
The code applies to them. In writing, in the sponsorship agreement and in the speaker confirmation.
Booth staff are attendees and are frequently the only people at the event who have not read anything you published. A line in the contract and a sentence at the sponsor briefing is the whole intervention, and it is the one that makes the code enforceable against the party with the most leverage.
What a code does when nothing happens
Most events have no incidents, and the code still did something.
It tells people what kind of event this is before they decide whether to come. A conference with a visible, specific, contactable code attracts a different and broader set of attendees than one without, and that effect is larger than the enforcement one.
That is worth saying because organisers who have never had an incident sometimes conclude the code was unnecessary. The absence of incidents is partly what it produced.
The short version
- Copy the text from an established conference and release yours under a free licence; ours came from PyCon UK 2015 and was reused elsewhere
- Adapt contacts, reporting mechanism and jurisdiction; do not soften the list of unacceptable behaviour
- What cannot be copied: named contactable people, a second contact, a decision procedure agreed in advance, and willingness to act
- Decide before the event who can eject someone, the range of responses, and what happens if the person is a speaker, sponsor or organiser
- We had no written decision procedure, which only matters under stress — which is when it arises
- A code changes who attends, so the absence of incidents is partly what it produced