The Wrong Kind of Safe

A foundation kept holding community engagement sessions to inform a new program, and the same people kept showing up in the same way. Some of the reasons were structural. Meetings were held during the day, and no childcare was offered. Some of the reasons were about power. The foundation was a large presence in a small community, and the people who came to the sessions were often the ones chasing grants. Fixing the structural barriers helped, but it didn’t really change the room's composition.

What did help was a single request. The organizers made a list of people they already knew and asked each to personally invite one person who was different from them and had never attended a foundation event. They didn’t define “different.” They only asked that the newcomer be, well, new.

This is Chris Corrigan’s story from a recent post on inviting diversity.1 The part that resonates with me is his diagnosis of why the obvious fix fails. Mandating a quota of diverse bodies, he argues, “is a technical solution to a complex problem.” It treats representation as a checkbox. Removing barriers was necessary but insufficient. The approach that worked was to lean on the trust that characterized pre-existing relationships. Strangers trusted the foundation enough to attend a session because a friend vouched for it.

The sociologist Ron Burt provides us with a vocabulary for what was happening here.2 The foundation’s network was dense and redundant. Its “usual suspects” inhabited the same social cluster, knew roughly the same things, and had already shared most of what they could. Between that cluster and the rest of the community sits what Burt calls structural holes, gaps in the network. His central finding is that novelty lives in those gaps. People whose relationships span structural holes encounter non-redundant information earlier and more often, and they disproportionately produce ideas that the rest of the network considers to be good. It’s not because they’re smarter, but because they’re standing at an intersection of different ways of thinking.

The “inviting diversity” solution described by Corrigan is a strategy for bridging structural holes. Each trusted insider is asked to reach across a gap and pull someone in from the other side. This personal trust is, at least in a limited sense, transitive. The new person doesn’t have to trust the foundation; they only have to trust their friend.

Scott Page argues that the case for the “diversity bonus” rests not on identity but on cognitive diversity.3 Cognitive diversity reflects genuinely different access to information, heuristics, representations, and mental models. In this context, identity diversity matters insofar as it reflects different tools and ways of thinking. It often does, but not always. The diversity bonus, therefore, is conditional twice over. The different ways of thinking must be present, and the group's culture must allow them to be heard and used. A group can be diverse on paper and capture none of the bonus if the people who think differently are ignored.

Most of what I’ve written here has focused on the second condition. I’ve published essays on why a well-composed group still converges on what everyone already knew and on how a room’s effective intelligence is bounded by what people are willing to say out loud. All of that takes the people in the room as a given and addresses the dynamics inside it.

Corrigan points at the first condition, which is upstream of everything I’ve been writing about. That is, the effective diversity has to be in the room, or at least within reach, before any amount of careful facilitation can surface its ideas. You can remove every social pressure, preserve every weak signal, and run a flawless process, yet still miss the perspective that would have changed the decision because the person who held it was never invited to speak. A quiet room and an empty one sound identical.

Here’s where I run into a dilemma. Our Reflect & Engage method works by subtraction. It removes the social interference around a meeting (attribution, status, time pressure) by having people reflect anonymously and asynchronously beforehand. It reduces the tendency for divergent views and minority perspectives to be suppressed in the moment. The method is designed to make candor cheap, on the theory that courage is a tax that good structure can subsidize.

So if the obstacle is that people at the edge of the network won’t walk into a room and be a visibly different newcomer, why not reach them through a channel that asks nothing of the kind? An anonymous, low-stakes prompt might gather genuinely divergent input from across structural holes without requiring anyone to perform in front of a group. As much as I have an incentive to advocate such an approach, there’s a problem.

Corrigan’s mechanism is the named friend who vouches for the foundation and extends an invitation, which the newcomer takes as a gesture of welcome and could be the start of a new and lasting relationship. Anonymity, on the other hand, strips out the relationship and the attribution that comes with it. That trade lowers the cost of candor, but it forfeits the very thing that can turn guests into trusted contributors.

Suppose you’re able to reach across the structural holes, lower the cost of speaking candidly, and deliver genuinely novel perspectives. The insider group still has to be able to receive them, and outside perspectives may be the input a group is best equipped to ignore.

The insider who holds unique information already struggles to find a warrant to share it. The outsider has even less standing. They have no history with the decision and no accountability for the outcome. In other words, they have no skin in the game.4 Anonymity means a perspective with no identity attached, which is the easiest of all to discount. Sometimes, a group is right to weigh unaccountable, context-free input with caution. The diversity bonus requires divergent views to be heard and used, and a faceless perspective from outside the room is likely to be difficult for the group to use.

A good process can put a novel perspective on the table. It can make ignoring it a deliberate choice rather than an accident. What it can’t do is make a group want what the outsider sees.

Corrigan’s generous broker, the trusted person who extends an invitation to a newcomer, brings diversity into the room. They do so by vouching for the organization. At the same time, the intermediary vouches for the newcomer, making the group more likely to listen to what the newcomer has to say. Anonymity can’t do that.

So the underlying question isn’t only how to reach people at the edge of the network, or how to make it safe enough for them to speak. It’s whether a perspective can be both safe enough to share and trusted enough to be used, at the same time, through the same channel. The relational design and the structural one each seem to buy one of those by spending the other. I don’t know whether you can have both, or whether wanting both is just asking for a smaller network.