Conflict, Repair, and the Long Table

Volume VI — Governance and the Commons

Nine movements, one table.

THE PLATE

A woman in linen standing at a doorway beside a young olive tree, warm light in the room behind her. — Plate VI.09Thursday, Noon, at the Cathedral Door.A conflict system is not a document either. It is a place, a time, and a group of people who are known to be there — and the cheapest of the three to provide is the place.

THE LETTER

You are writing rules for something — a cooperative, a commons, a firm, a community land trust, a team of eleven — and you have reached the part that everyone reaches and nobody enjoys. What happens when two members cannot agree? What happens when one of them takes more than their share? What happens when somebody simply will not stop?

Most people writing that clause do one of two things. They write the parties shall seek to resolve the matter amicably, which is not a mechanism, or they write a termination clause, which is a mechanism with exactly one setting. Both are expensive, and the second is far more expensive than the first, and almost nobody has ever put a number on either.

This chapter puts the numbers on. What a lawsuit actually costs and how long it actually takes, from the courts' own published statistics rather than from anybody's estimate. What mediation settles and what it does not. What restorative processes do to repeat offending, in effect sizes from randomised trials, including the trials where the effect was zero and the one where it went the wrong way. And then the thing this house has not seen written down anywhere: what a graduated sanction system costs to run, per member, per year — and the dispute rate above which running one is cheaper than throwing people out.

That ratio is the instrument. It is a single number and you can compute it for your own organisation this afternoon, before you write a word of the clause. It will tell you something that will probably surprise you, and the surprise is the most useful thing in this chapter.

One promise about the honest part. There is a real vulnerability in every consensual process, and it is not a small one: a repair process needs a party willing to be accountable, and a system with no fallback to coercion can be taken apart by anybody who simply refuses. We will price that refusal, in percentage points from a published compliance study, and then price the fallback that answers it — and the most interesting finding in the chapter is what the fallback does to the ninety-eight per cent of cases that never reach it.

— The Editors

DISCOVERY

What is already working

Thursday, noon, Valencia. The Tribunal de las Aguas — the Water Court of the plain of Valencia — sits at the Apostles' Door of the cathedral, outdoors, in public, and has done so for approximately a thousand years. Eight judges, the síndicos, are elected by the irrigators of the eight main canals from among the irrigators themselves. They are farmers. They are unpaid. Procedure is oral: the complainant speaks, the respondent answers, the tribunal confers and rules, usually within minutes. There are no lawyers, no written pleadings, no transcript, and no appeal. Fines are in the local unit and are modest. Fifty- two sittings a year for a thousand years is on the order of fifty thousand hearings, and the institution's operating cost is a set of chairs.

Elinor Ostrom put the huerta of Valencia at the centre of Governing the Commons for a specific reason. It is not that the Water Court is picturesque. It is that it satisfies two of her design principles at once, cheaply, and has kept satisfying them across a millennium of regime change: principle four, monitoring by monitors who are accountable to the appropriators or are the appropriators themselves — the síndicos are irrigators and stand for election among irrigators — and principle six, rapid access to low-cost local arenas for resolving conflicts. Thursday. Noon. The cathedral door. You do not need to know anything else to use it.

Ostrom's eight, and what happened when somebody checked them. In 2010, Michael Cox, Gwen Arnold and Sergio Villamayor-Tomás re-analysed ninety-one studies of community-based natural resource management against Ostrom's principles and published the result in Ecology and Society. The principles held up. They reformulated the eight into eleven for precision, and among the best-supported were exactly the two this chapter is about: monitoring, and graduated sanctions. That is an unusual thing to be able to say about a governance framework twenty years after it was proposed, and it is why the arithmetic later in this chapter is worth doing carefully — you are building on something that has been checked.

Mondragon, and the forum nobody photographs. The Basque federation runs on the order of ninety-five cooperatives employing about seventy thousand people — a mean of roughly seven hundred and thirty-seven people per cooperative. What is less often described than its ownership structure is its conflict structure: each cooperative has a Social Council, distinct from the governing council and the management, whose standing business is the things that in a conventional firm arrive as grievances. It is a forum that exists before it is needed. That is the whole trick, and it is the same trick as the cathedral door.

Restorative conferencing, and the trials that tested it. Beginning in the 1990s, Lawrence Sherman and Heather Strang ran a programme of genuinely randomised controlled trials of restorative justice conferencing — in Canberra (the RISE experiments) and then in the United Kingdom through the Justice Research Consortium. Offenders who had admitted responsibility were randomly assigned either to the ordinary court process or to a facilitated conference with their victim. This is rare and valuable: most of what is claimed about conflict processes is claimed from case studies. Here there is randomisation, and therefore a defensible causal claim, and therefore an effect size we can put in a table.

And the one nobody expects. Between 1975 and about 1980, Maine's small claims courts ran mediation alongside adjudication, and Craig McEwen and Richard Maiman went and counted what happened afterwards — not what people said in the room, but whether the money actually moved. That study is forty years old and it is still the cleanest natural experiment on compliance anybody has. Its central pair of numbers is in the Arithmetic below, and it is the reason this chapter can price a refusal rather than merely deplore one.

Five cases, one pattern: in each, a forum existed before the dispute did, and its existence was cheaper than its absence. Nobody in Valencia decides on a Tuesday whether to have a water court. It is Thursday and the court is there.

THE ARITHMETIC

What works, what does not, and where the line sits

First, the price of the route you already have.

Most organisations have precisely one written dispute mechanism, and it is a court. So price it. Not from estimates — from the courts' own statistics and from the federal research body that surveys the lawyers who ran the cases.

  route                        elapsed            cost
  -------------------------------------------------------------------
  civil trial, US district     ~30 months      $15,000 plaintiff
    (913 days)                                 $20,000 defendant
  employment arbitration        361.5 days      employee wins 21.4%
  commercial mediation          1 day           settles 72% same day
  internal panel                ~21 days        £330 per hearing

The thirty months is the Administrative Office of the U.S. Courts' own reporting of median time from filing to trial in civil cases in the district courts — the figure moves through the high twenties and high thirties of months across recent years, and thirty is the round number to hold. Converted at 30.44 days to the mean calendar month, that is 913 days. Against a commercial mediation that occupies one day, the duration ratio is 913×. Against an internal panel that takes three weeks from complaint to finding, it is 43.5×.

The $15,000 and $20,000 are medians of reported litigation costs from the Federal Judicial Center's National Case-Based Civil Rules Survey (Lee and Willging, 2010), and they are lower than most people expect for a reason worth saying plainly: they are medians across closed cases, the overwhelming majority of which settled without ever reaching trial. That is the cost of the route, not the cost of a trial. As a share of what was at stake, the medians were 1.6 per cent for plaintiffs and 3.3 per cent for defendants — the defendant pays about 1.33× what the plaintiff pays to obtain the same outcome, because the defendant cannot choose not to be there.

And the system-level figure, which is the one that should be read twice. RAND's study of the American tort system (Kakalik and Pace, 1986) found that of every dollar the system spent, about 46 cents reached the injured party. Fifty- four cents of every dollar is the cost of finding out who was right.

Second, the middle rungs, measured rather than advertised.

Arbitration is sold on speed and it does deliver speed. Alexander Colvin's study of American Arbitration Association employment cases (2011) found a mean 361.5 days to disposition — 11.9 months, or about 18.1 months faster than the trial track. But the same study found the employee win rate in arbitration was 21.4 per cent, against 36.4 per cent at federal court trial — a ratio of 1.70× in the court's favour — with a median award of $36,500 where the employee won.

Read those two findings together and you have the honest sentence about arbitration: it is faster, and for the weaker party it is worse. Speed is not free and it is not neutral. A conflict system designed by the party that will usually be the respondent will produce exactly this shape, and noticing that is part of designing one that does not.

Mediation's record is stronger and its denominator is narrower. The CEDR Mediation Audit, run biennially across the UK commercial mediation market, reports roughly 72 per cent of cases settling on the day of the mediation and a further 20 per cent shortly afterwards — an aggregate of about 92 per cent, with 8 per cent not settling. Hold the denominator: those are cases that reached a mediation, which means both parties agreed to be in the room.

Third, restorative justice, with the effect size the trials actually produced.

The Campbell systematic review (Strang, Sherman, Mayo-Wilson, Woods and Ariel, 2013) pooled ten randomised trials of restorative justice conferencing and reported a standardised mean difference on repeat offending of about d = 0.155, favouring conferencing.

That is a small effect and it should be stated as one. But a standardised mean difference is not a quantity anybody can act on, so convert it. Using the standard logistic conversion — multiply d by π/√3 = 1.8138 to get a log odds ratio — gives an odds ratio of 1.3246. Applied to a two-year reconviction base rate of 50 per cent, that moves the treated rate to 43.0 per cent: an absolute reduction of 7.0 percentage points, and a number needed to treat of 14.3. Run fourteen conferences and, on this evidence, one person who would have reoffended does not.

Now cost that, conservatively and from the bottom, rather than repeating a headline. At an illustrative £1,500 per facilitated conference, preventing one repeat offender costs £21,482. One prison place in England and Wales costs on the order of £48,000 a year. That is a benefit-to-cost ratio of 2.23 : 1 on avoided custody alone. The published figures run from about 3 : 1 to 14 : 1 because they count police time, court time, victim costs and subsequent offences over a longer horizon; ours is lower because it counts one year of one prison place and nothing else. We give the conservative reconstruction and let the reader see the gap — a ratio you can rebuild from its parts is worth more than a larger one you cannot.

Fourth — and this is the honest negative, and it is a real one — where this loses.

Restorative processes require a party willing to be accountable. Every trial above randomised only cases where the offender had already admitted responsibility and both parties consented. That is not a footnote about methodology. It is a statement about the population the evidence covers, and it excludes exactly the person a conflict system most needs to handle: the one who denies everything and will not come.

And the trials are not uniformly positive, which is what makes them believable. In the Canberra RISE experiments, the drink-driving arm showed no reduction in reoffending and a small increase in the conference group; and in one violent- crime arm, offending among Aboriginal offenders rose. Sherman and Strang have reported these as consistently as the positives. The effect is real, it is small, it is stronger for violent crime than for property crime, and in at least two named populations it went the other way.

So price the refusal. McEwen and Maiman's Maine data is the place it can be priced, because they followed the money. Full compliance with mediated agreements was 70.6 per cent. Full compliance with adjudicated judgments was 33.8 per cent. Consent roughly doubled voluntary payment — which is the pro-mediation finding everyone quotes.

The finding nobody quotes is the other end of it. Non-compliance with a mediated agreement is 29.4 per cent, and a mediated agreement has nothing behind it. A court order, complied with voluntarily only 33.8 per cent of the time, leaves 66.2 per cent outstanding — but that residual can be enforced. So there is a break-even in enforcement effectiveness, and it is computable:

  e*  =  (0.706 - 0.338) / (1 - 0.338)  =  0.368 / 0.662  =  55.6%

Consent alone beats a court order unless enforcement recovers more than about 56 per cent of the unpaid balance. Below that, the voluntary route delivers more actual compliance than the coercive one. Above it, the order wins. That single figure is the whole argument about fallbacks, and it means the question is not philosophical. It is: how good is enforcement where you are?

Fifth: the instrument. What a graduated system costs, and the rate at which it beats exclusion.

Ostrom's principle five is graduated sanctions — a ladder whose first rung is a word and whose last rung is exclusion, with real rungs in between. Price the ladder for a cooperative of 200 members.

Standing cost, whether or not anything happens. Training three panellists two days a year: £1,500. Panel time held in reserve, three people at twelve hours each: £1,080. Convening and keeping the record: £920. F = £3,500 a year.

Per-dispute cost. Three panellists for three hours: £270. Notice, minute and follow-up: £60. c = £330 per hearing.

At an observed dispute rate of λ = 0.05 per member per year — ten disputes a year in a house of two hundred — the arithmetic is:

  F / N       =  £3,500 / 200        =  £17.50  per member per year
  λ · c       =  0.05 × £330         =  £16.50  per member per year
  ------------------------------------------------------------------
  total                                 £34.00  per member per year
                                       (£2.83  per member per month)
  whole system                        £6,800   per year

Thirty-four pounds per member per year. That is the answer to the question this chapter was written to answer, and it is smaller than most people guess by about an order of magnitude.

Now the other side. What does not having the ladder cost? It costs exclusions. Price one. Replacement cost of a member on £30,000 at the 21 per cent of annual salary that Boushey and Glynn found across thirty case studies is £6,300. Add a 15 per cent chance the excluded member brings a claim costing £8,500 to defend: £1,275 expected. X = £7,575 per exclusion — 23.0× the cost of one internal hearing.

Let q be the share of disputes that end in somebody leaving when there is no forum, and s the share that still reach exclusion when there is one. The ladder is cheaper when:

  F/N  +  λ·c  +  λ·s·X   <   λ·q·X

                F / N
  λ*  =  ------------------------
           (q - s)·X  -  c

With q = 0.30 and s = 0.05, the denominator is (0.25 × £7,575) − £330 = £1,563.75, and:

  λ*  =  £17.50 / £1,563.75  =  0.01119 disputes per member per year
      =  1.12 disputes per hundred members per year
      =  2.24 disputes per year in a house of two hundred

Here is the cut. Notice what happens when you change the size of the house. At fifty members λ* is 0.04476; at five thousand it is 0.00045. In every case the threshold in whole disputes is the same: 2.24 a year. Because the standing cost is fixed, the break-even is not a rate at all — it is an absolute count of about two and a quarter disputes a year, for a commune of fifty or a company of five thousand.

Two disputes a year. That is roughly one every five and a half months. There is no organisation of humans on record that is below that line. The observed rate in our two-hundred-member house is 4.5× the threshold, and that house is a quiet one.

So the finding is not a conflict system pays for itself in large organisations. It is: you already own a conflict system, you are paying for it in exclusions, and it is the most expensive model available. The decision was never whether to buy one. It was made the day you wrote a termination clause and nothing above it.

Two honest limits on that, because a threshold this dramatic deserves them. First, the whole result rests on q − s, the share of disputes the ladder diverts from exclusion, and that is the chapter's most load-bearing assumption — it is the number a reader should replace first with their own. Second, there is a floor: the ladder must divert at least 4.4 per cent of disputes (that is c/X), or the hearings cost more than the exclusions they prevent and no threshold exists at all. A forum that hears everything and changes nothing is not cheap. It is just a slower way of arriving at the same door.

DREAM

What becomes ordinary

In the organisation that has built this, the conflict forum is boring, and its boringness is the achievement.

It meets on a known day whether or not there is business, the way the Water Court meets on Thursday. Most weeks there is nothing and the meeting lasts four minutes. That is not waste; it is the standing cost, it is seventeen pounds fifty a head a year, and it is what makes the forum available on the week it is needed without anybody having to be brave enough to convene it. Nobody ever has to propose that there be a hearing. The hearing already exists; the only question is whether anything is brought to it.

The panel is elected from the membership and sits for a fixed term, and serving on it is understood as ordinary work rather than as a promotion or a punishment. People who have served on it are noticeably better at the rest of their jobs, which nobody planned and everybody has noticed.

The ladder has real rungs and they are written down: a conversation, a recorded conversation, a facilitated conversation, a finding with conditions, a suspension of a specific right, and — at the top, rarely, and never quickly — exclusion. Because the rungs exist, the early ones get used. In an organisation whose only sanction is dismissal, the first four grievances are handled by pretending they are not happening, because the only available response is disproportionate to all of them.

The record is kept and it is readable. Not to punish anybody with their history, but because the fourth time the same two departments collide, somebody can see that it is the fourth time, and the finding can be about the interface rather than about the people standing in it. Monitoring, in Ostrom's sense, is not surveillance of members. It is the organisation being able to see its own patterns — and the monitors are the members, which is the entire content of principle four.

And there is a last rung that is real and almost never used: a named external arbitrator, on a standing retainer, whose decision binds. It costs about nine pounds fifty per member per year. It is invoked roughly once every five years. Its function is not to decide those cases. Its function is that everybody in every conversation on every lower rung knows it is there.

DESIGN

The structure that gets there

The ladder, with its rungs named.

A graduated sanction system is six rungs and a published rule about how one moves between them. The rungs:

A word. Any member to any member, unrecorded, no process.
A recorded word. A short note filed by either party. No finding, no consequence; it exists so that a pattern can later be seen.
A facilitated conversation. One trained facilitator, one hour, no panel. This rung handles most of everything.
A panel hearing with a finding. Three panellists, three hours, a written finding with conditions and a review date. £330.
A suspension of a specific right — from a rota, a budget, a decision forum — for a stated period, with restoration automatic on completion.
Exclusion, and only after five.

The rule for moving is that a rung is used twice before the next is available, and that nobody skips except for defined conduct written into the membership agreement. Both halves matter. Without the first, the ladder collapses to its top rung under pressure. Without the second, the ladder becomes a shield.

The forum, built to Ostrom's sixth principle.

Low cost means four things, and all four are design decisions, not aspirations. Fast — a hearing within twenty-one days of a complaint, which is 43.5× faster than the trial track. Local — heard by people who know the work, which is what makes a twenty-minute hearing possible where a court would need two days of evidence. Cheap — no representation, because the moment one side brings an advocate the other must, and you have rebuilt the expensive thing inside the cheap one. Public within the membership — the record readable by members, because a forum nobody can see cannot be trusted and a forum everybody can see does not need to be.

The monitors, built to the fourth principle.

Monitoring fails in two directions and the design has to refuse both. Monitors appointed from above produce information the membership does not believe. Monitors with no accountability produce none at all. Ostrom's formulation is precise: monitors are accountable to the appropriators, or are the appropriators. So: the panel is elected by the membership, sits a fixed term with a term limit, and publishes an annual count — how many complaints, at which rungs, how many reached which outcome, and median elapsed days. That annual count is the only monitoring apparatus required, and it is the thing that lets you compute your own λ and compare it against your own λ*.

The fallback, and the shadow it casts.

Write a binding arbitration clause into the membership agreement, name the arbitrator or the panel from which one is drawn, and put them on a small standing retainer. At £1,200 a year plus £3,500 per hearing actually held, with 2 per cent of disputes reaching it, the expected cost is £1,900 a year — £9.50 per member per year, a 27.9 per cent uplift on the £34.00 system, taking the whole thing to £43.50 per member per year, or £3.62 a month.

What that £9.50 buys is not the hearings. It is the compliance in the ninety- eight per cent of disputes that never reach it. This is Mnookin and Kornhauser's bargaining in the shadow of the law (1979): parties negotiating privately settle against their estimate of what would happen if they did not, so the terminal rung sets the price of every rung beneath it. Change the terminal rung and every conversation on the ladder changes, including the ones where nobody mentions it.

And it is straightforwardly cheap. Non-compliance without a fallback runs at 29.4 per cent; ten disputes a year gives 2.9 broken agreements, worth £22,271 a year if each becomes an exclusion — £111.35 per member per year. The fallback costs £9.50. It pays for itself if it prevents a quarter of one exclusion a year.

DESTINY

How it holds when nobody is pushing

Three things keep a conflict system alive, and they are not the three people expect.

It meets on a schedule, not on demand. A forum that convenes only when somebody requests it requires somebody to be brave, and bravery is not a renewable input. A forum that meets on the first Thursday requires nothing of anybody. This is why the Water Court has a day and not a procedure for getting one.

Its count is published. One page a year: complaints, rungs, outcomes, median days. It costs almost nothing and it does two jobs — it lets the membership see that the forum is being used, and it produces the λ that goes into the threshold calculation next year. A forum that publishes no count is indistinguishable from a forum that has quietly stopped.

Somebody who is not the founder has run one. One person who knows how to chair a hearing is a bottleneck with a name. Three is an institution.

Now the failure modes, named as honestly as we can.

Capture. A panel elected by a majority faction becomes an instrument of that faction, and it is worse than having no panel because it carries legitimacy. Term limits, rotation, and a recusal rule with teeth are the available defences, and they are not complete defences.

Drift to the top rung. Under stress an organisation stops using rungs three and four because they take three weeks, and reaches for six because it takes an afternoon. The tell is in your own published count: if the ratio of rung-three to rung-six outcomes is falling, the ladder is collapsing, whatever anybody says at the meeting.

The refuser who is correct. Sometimes the person who will not engage is right that the forum is not fair to them — a minority member, a whistleblower, someone raising a complaint against the panel's own constituency. A system that treats every refusal as bad faith will crush the one refusal that was the most important signal it ever received. The available defence is that the fallback runs both ways: any member can escalate to the external arbitrator, not only the organisation.

And the plain one: it fails when there is nothing behind it. A repair process with no terminal rung is exploitable by anybody patient enough to keep refusing, and the exploitation rate is not hypothetical — it is 29.4 per cent, measured, forty years ago, in Maine. Design the last rung so that you never have to use it. Do not design it away.

DELIGHT

What it feels like

There is a specific relief, and it arrives before anything is resolved. It is the moment a person with a grievance learns that there is a Thursday. Not that they will win — they may not — but that the thing they are carrying has somewhere to go, on a known date, without them having to persuade anyone that it deserves a meeting. The carrying is most of the weight, and the calendar takes it off.

Then there is the pleasure of a short hearing. Twenty minutes, standing up, two people who know the work explaining it to three people who also know the work, and a finding by lunch. Anybody who has sat through a two-day process about a three-hour problem knows exactly what is being described. Speed is not a compromise on justice here. It is a form of respect — for the facts, which are usually simple, and for the people, who have work to do.

And the best one, which you will only notice a year in: the arguments get better. When people know there is a forum, they start bringing the actual complaint rather than the deniable version of it, because the deniable version does not survive being said out loud on a Thursday. An organisation with a working conflict system is not an organisation with less conflict. It is one where the conflict is about what it is actually about, and that turns out to be a much more pleasant place to work.

OPERATIONALIZE THIS

At the level of finance

The instrument: a Repair Reserve with a graduated-sanction ladder and an arbitration backstop.

This is a provision, a published procedure, and a retainer. It is small, it sits on the balance sheet in a place your auditor already understands, and it is presented to a board not as an ethical proposal but as a substitution of a cheap cost for an expensive one that is already being incurred.

The mechanics.

Reserve size. Twelve months of expected cost: F + (λ · N · c) + the arbitration retainer and expected hearings. In the worked house that is £3,500 + £3,300 + £1,900 = £8,700 a year, which is £43.50 per member per year. Fund it annually from operating expenditure; it is too small to capitalise and pretending otherwise invites a discussion you do not need.
Balance-sheet treatment. The standing cost is an operating expense. The arbitration retainer is a prepayment amortised over its term. Where the membership agreement creates a present obligation to hold a hearing on request, treat unresolved live cases as a provision under your ordinary provisions policy and disclose the count, not the names. Talk to your auditor once, early; this is a conversation about provisions and contingent liabilities, which is a conversation they have every year.
The counterparty. One named external arbitrator or a named panel — a CEDR- accredited mediator, an AAA or JAMS panellist, a sector body's scheme — on a standing retainer, with a written undertaking that either party may invoke. Both-ways invocation is the term that makes it legitimate rather than managerial, and it costs nothing to include.
The monitor. Internal audit, or the secretary, publishes the annual count: complaints, rung, outcome, median elapsed days, and the resulting λ. Ostrom's principle four, discharged in one page a year.
The term. Three years, with the whole procedure reviewed against the published count at the end of year one. Review against a number, not against a mood.

The number that decides it. One figure, on the front page of the paper:

                    F / N
   λ*  =  --------------------------        and then:      λ  ÷  λ*
            (q - s) · X   -   c

Where λ is your own observed disputes per member per year and λ is the threshold. If that ratio exceeds 1, the ladder is already cheaper than what you are doing. In the worked house it is 4.5×. Compute yours with your own head count, your own replacement cost, and your own honest estimate of how many disputes currently end in somebody leaving. The estimate of q* is where the argument will be, so make it in the room, out loud, with the people who would know.

The first ninety days.

Day	Action	Artifact
1–15	Count last three years' exits and classify: resignation, dismissal, unresolved dispute	The exit classification — this produces q
16–30	Compute X from your own payroll and your own claim history	A one-page cost of exclusion
31–45	Compute λ* and λ. Put both on one page	The threshold page
46–60	Write the six rungs and the movement rule into the membership agreement	Draft procedure
61–75	Elect the panel. Train them. Name the arbitrator and sign the retainer	Panel roster, signed retainer
76–90	Publish the day, the rungs and the count template. Hold the first meeting with no business on the agenda	The first Thursday

That last line is the one to keep. The first sitting should have nothing on it. A forum that first appears on the day it is needed is a tribunal. A forum that has already met four times about nothing is furniture, and furniture is what you want.

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

Think of a serious disagreement here in the last two years that ended well. Who was in the room, how long did it take, and what did they do that worked?
Where do people already take things that are bothering them — the actual place, not the official one? What does that informal forum have that our formal one does not?
When has somebody in this organisation apologised properly and it changed something? What made that possible for them?

Dream — what becomes possible

If we had a standing day when anything could be raised, what would come up in the first three months that has been carried for years?
Imagine a year from now we publish one page of numbers about how we handle disagreement, and we are proud of it. What is on that page?
If every person here knew exactly what the second rung and the third rung were, what conversation would happen earlier than it does now?

Design — what we build

What are the real rungs between a quiet word and you are out, and which of them do we already use without having named them?
Who would we all trust to sit on this — and what would we have to change so that serving is ordinary work rather than a burden or a promotion?
What is the last rung, who can reach for it, and are we willing to let a member reach for it against us as well as the other way round?

Destiny — how it holds

What number would we watch to know whether the ladder is quietly collapsing to its top rung, and who would see it first?
What would have to be true for this to still be sitting on a Thursday when everybody currently in this room has gone?
If somebody refuses to take part, what do we want to be true about how we treat them — and how do we make sure the one refusal that was right still gets heard?

WORKS CITED

Boushey, H. and Glynn, S. J. (2012). There Are Significant Business Costs to Replacing Employees. Center for American Progress.

Centre for Effective Dispute Resolution. The CEDR Mediation Audit. Successive biennial editions, London.

Colvin, A. J. S. (2011). "An Empirical Study of Employment Arbitration: Case Outcomes and Processing Time." Journal of Empirical Legal Studies, 8(1), 1–23.

Cooperrider, D. L. and Whitney, D. (2005). Appreciative Inquiry: A Positive Revolution in Change. Berrett-Koehler.

Cox, M., Arnold, G. and Villamayor-Tomás, S. (2010). "A Review of Design Principles for Community-based Natural Resource Management." Ecology and Society, 15(4), 38.

Kakalik, J. S. and Pace, N. M. (1986). Costs and Compensation Paid in Tort Litigation. RAND Institute for Civil Justice, R-3391-ICJ.

Lee, E. G. and Willging, T. E. (2010). Litigation Costs in Civil Cases: Multivariate Analysis. Federal Judicial Center, National Case-Based Civil Rules Survey.

McEwen, C. A. and Maiman, R. J. (1984). "Mediation in Small Claims Court: Achieving Compliance Through Consent." Law & Society Review, 18(1), 11–49.

Ministry of Justice (England and Wales). Costs per Place and Costs per Prisoner. Successive annual editions.

Mnookin, R. H. and Kornhauser, L. (1979). "Bargaining in the Shadow of the Law: The Case of Divorce." Yale Law Journal, 88(5), 950–997.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Shapland, J., Atkinson, A., Atkinson, H., Dignan, J., Edwards, L., Hibbert, J., Howes, M., Johnstone, J., Robinson, G. and Sorsby, A. (2008). Does Restorative Justice Affect Reconviction? The Fourth Report from the Evaluation of Three Schemes. Ministry of Justice Research Series 10/08.

Sherman, L. W. and Strang, H. (2007). Restorative Justice: The Evidence. The Smith Institute, London.

Strang, H., Sherman, L. W., Mayo-Wilson, E., Woods, D. and Ariel, B. (2013). "Restorative Justice Conferencing (RJC) Using Face-to-Face Meetings of Offenders and Victims: Effects on Offender Recidivism and Victim Satisfaction. A Systematic Review." Campbell Systematic Reviews, 9(1).

United States, Administrative Office of the U.S. Courts. Judicial Business of the United States Courts, Table C-5: median time intervals in civil cases, U.S. District Courts. Successive annual editions.

Whyte, W. F. and Whyte, K. K. (1991). Making Mondragón: The Growth and Dynamics of the Worker Cooperative Complex, 2nd edn. ILR Press.

Note on figures. Every figure in this chapter is computed in lib/verify/VI_09.py and printed with its inputs, its intermediate terms and its source. Figures taken from published research are marked CITED there; scenario inputs — the two-hundred-member house, the panel budget, the escalation shares q and s — are marked ILLUSTRATIVE and are not presented as findings. The restorative-justice benefit-to-cost ratio is rebuilt conservatively from the effect size and one year of custody, which is why it is lower than the published range; the gap is stated rather than smoothed.