On Group Discussion Formats

I’ve been thinking about group discussions, particular those where a group of four or more people are exploring and learning about a complex issue together. I’m doing this both because I want to reboot the Futures Salon I’ve been running with Arunabh and because I’m currently gearing up to run an unconference for the Seattle XR community. I’m particularly interested in what I can do to facilitate better discussions in those contexts.

In this post, I’m going to summarize some of the formats I’ve considered, both to help clarify my thoughts and so I have a resource to point at later. Before getting into the formats, I want to touch on the goals of such discussions and talk about why they often aren’t met.

To my mind, then, the goals of this kind of group discussion are a mix of sharing knowledge and ideas to educate each other and developing new knowledge and ideas by exploring existing ones through friendly argument and research. Sounds easy, right?

The problem is that groups have a tendency to degrade into dyads with an audience, where two people dominate the conversation, possibly with another person or two throwing in extra remarks from time to time. Though the dyad can shift over time, they tend to result in discussions where a small number of powerful or loud personalities dominate. This in turn results in less diverse input into the discussion, less opportunity to critique ideas from different perspectives and, in the worst case, to push away certain members of the group. Making this harder, a lot of people just aren’t very confident putting their ideas out there, and though it can be a lot of work to help people speak freely, it would be a shame not to find a way to include them.

It’s worth noting that this pattern is not necessarily due to malice or aggression on anyone’s part – some people are just more comfortable talking than others, and some people are just louder and bolder while others are more hesitant and careful. There are, of course, toxic behaviors that make things worse – chronic interrupters, shutdown artists, and violent communicators to name a few – but even in the absence of these, the problem persists. Let’s begin with the assumption that everyone is there with good intent and that you’re already prepared to call people on those behaviors if they crop up.

With that done, fixing these issues still isn’t easy. I’ve learned from running retrospectives and facilitating meetings that it is possible, but requires leadership and creativity to set up structures so everyone can participate. To be clear – I’m not holding myself out as an expert on any of this, merely as a traveler on the long road towards improvement who feels like writing a guidebook.

So, without further ado, here are some discussion structures that I think are interesting:

• Show & Tell. A topic is announced in advance. All members are encouraged to do a little bit of reading or thinking beforehand and are asked to bring something to the group that they will present in a discussion section that they lead. One individual may lead the
  discussion overall, but their job is primarily to facilitate the transitions between each member’s chance to talk. Advantages of this method are that everyone gets a chance to speak when presenting their item and can be presumed to be a significant part of the discussion that follows, and participants are more likely to be invested in the topic as a result of having to prepare something to show. Unfortunately, though it gives everyone a chance to present their item, it does little to help people participate outside of this, and strong personalities could easily dominate the discussion around each item. Another limitation is that this structure doesn’t scale terribly well.
• The fish bowl – A subset of people are designated as speakers and seated centrally, while everyone else sits around them as an audience. Conversation proceeds between the speakers and can follow a topic or be entirely free-form. Some mechanism is defined whereby members of the audience switch out with speakers. Variations include whether speakers choose when to switch, whether they choose who to switch with, whether audience members can request a switch, and whether there’s a time limit. An advantage of this format is that it supports fairly large groups – the audience could easily consist of twenty or more people without posing a problem, though the number of speakers should probably not exceed four. An obvious problem is that speakers might not want to switch out, so some mechanism to regulate this is needed. Outside of kicking things off, no moderator or facilitator is required for this format.
• The split party. Participants are divided into small groups of three to four members and each is given a topic or task to discuss. Topics should be different aspects of a larger issue, and need not be unique to each group. At some frequency, people move between groups and either continue the discussion or take up a new topic. Optionally, before each transition, each outgoing group may elect a speaker to summarize and share the points and conclusions reached in their discussion. This format is good for contentious or engaging topics where lots of people want to have a say. It can also be an effective way of grappling with a large issue that can be divided into parts. Lastly, it can be very effective in groups where the members don’t know each other well. To facilitate mixing, groups should be randomized.
• The long conversation. Participants are arranged into a list. Seats are arranged with two for speakers and the remainder facing them as the audience. Speakers one and two begin the discussion while the other speakers sit in the audience. After some interval, perhaps 10 minutes, speaker number one bows out, number three joins, and the discussion continues. This process proceeds with each participant speaking for two intervals. In the last interval, the first participant returns and joins the last participant to wrap up the discussion. This structure allows diverse participants to take the discussion in a variety of directions. Participant ordering could be randomized, or it could be curated to ensure interesting conversations. It also scales up well with a large audience of non-speaking participants and could, in principle, be split over multiple occasions.
• Question time. This method is particularly inspired by an agile retrospective. In the first part, participants generate questions to be discussed, either silently on their own or using one of the various brainstorming techniques that exist. The one rule is that no attempt is made to get to an answer for now. At some point, maybe one quarter to one third of the way into the event, the group stops generating questions. In some order, each participant either selects a question to explore or chooses to join another participant in exploring the question they selected. Groups have a maximum size, perhaps four, and are responsible for generating an a answer they’re ready to present. Once some period of time has passed, all groups are given a warning and shortly after, discussion is ended. Groups are then asked to present their answers. If time remains, the discussion and answering round is repeated with new groups. The chief benefit of this structure are its emphasis on reaching conclusions. Its chief limitation is that it imposes more structure than some people may be comfortable with. Many variations are possible, including the use of a moderator, the process for generating questions, and exactly how question groups are formed.

The goal of all of these protocols is to create a social environment that encourages many different people to have their say, where having that say is fun and non-confrontational, and where conversation is gently steered towards productive topics. They differ from one another chiefly in how many people can talk at once, what topics they are suitable for, and the extent to which they support an audience. All of them work to prevent one or some speakers from dominating, though some benefit from a moderator.

None of these protocols are perfect, but all seem likely to result in interesting conversation. They can (and should) be hybridized, and there are likely many other models that could be followed, but they provide a good starting point for organizing group discussions of the sort I like to participate in.