We Want Vibes
Everyone wants vibes. People want to feel like we're in this together. wgmi or ngmi - but at least we're not alone. It's a human tendency of wanting to belong.
Measuring vibes is hard. It's a feeling people have towards another group of person. On Farcaster, vibing is related to how you feel about a channel. Does a channel feel like it's yours, your post belong there, you know the people on it, it has become your digital third space.
Scientists have never thought about measuring vibing, the concept is too fizzy, ill-defined and open for interpretations. However, similar words have been used. Closure is a good example. It isn't exactly vibing, but a facet of it.
Closure combines two features: It is a sign of people knowing each other, what they do, and what they need, and willingness to spend energy and time coordination people. Everyone who is a professional cat-herder knows how hard it is to coordinate people.
More specifically, closure happens when you introduce two of your friends to each other and now you don't have to do this awkward coordination dance anymore and say stuff like "I saw the movie already with Becky". But coordinating activities with two people is more complex than with one person.
Eric Quintanea's (2013) studied the evolution of communication patterns in two teams and linked it to performance. Communication pattern are the turn-taking that happens between people. You talk, I answer, Bob interrupts. He and his colleagues used a short-term frame (24hrs) and a long-term frame (4 months).
In the study, only one team showed sign of closure. Looking for answers, he spoke with the manager of those teams and learned that the team without the more complex coordination mechanisms (closure) lacked team spirit. They also performed less well. There is a relation between closure (an easily observed communication pattern) and team vibes (a fuzzy hard to measure concept).
Does closure exists on Farcaster?
I'm making the following assumption
If complex interaction patterns aka closure is a structural sign of "vibes" then channels with more of these structures have more vibes.
Now that assumption is a bit crude and simplistic. Because data is just data and without the stories and faces behind the data points, we're only scratching the surface. At this moment, if you are craving stories, check out Outcaster excellent newsletter and tune into gmFarcaster.
To test my hypothesis, I needed to pick a couple of channels. But not just any channel. Some channels are like little cafes on Farcaster-land where people come together talking about a topic (e.g., hiking), other channels have been described as "old-school Facebook walls" (e.g., teddit, july), while another group of channels are more like customer support channels (e.g., MetaMask, supercast). And then of course there is Degen, Warpcast or other channels that feel more like huge shopping malls or casinos where everyone is looking for their personal gains.
I've picked four channels for this research I'm somewhat familiar with. I avoided too active or too big channels (too much noise, bigger datasets leading to longer computation times).
FarcastHer: No need to explain why. I'm active in this channel. Over the months the casters have come together and a vibe exists. I'm pretty certain of this.
Data: Data is an interesting channel. Part Q&A, part shill-your-work. The community is international and coming from different data angles.
Nouns: Nouns is an interesting use case as the community exists also beyond Farcaster under the same name and spirit. Some nouners know each other IRL. We know that a community exists, and that there should be (pocket) of vibes
Purple: Farcaster's DAOs, with overlapping members from nouns. Also full of OGs. Some met IRL at FarCon or other events.
I've taken all interactions in these four channels for the past 3 months. Casts that got no reactions (like, recast), or reply or mention nobody were deleted. In this first analysis, I'm not making a distinction between low-effort interactions (likes and recasts) and high-effort interactions (replies and mentions). All interaction patterns have been summarized per day. An interaction pattern is any like, recast, reply or mention between two casters (FIDx, FIDx+1).
The code to clean the data and run the analysis is available here.
Result 1: The Nouns channel got in my way making nice graphs
Result 2: Monologues > Dialogues
The bedrock of any relationship is a back-and-forth between people. One way to measure this is through reciprocity. Reciprocity is the technical term for "two people talking with each other". A reciprocal interaction exists when Anna mentions Betty and Betty replies to Anna. It's thanks to Betty's reply that reciprocity exists.
Another way reciprocity manifests itself in social networks is when Carla casts, Dora replies to the original cast, and Carla replies to Dora's reply. The most common form of reciprocity on a social network is when Erica casts, Francesca likes Erica's cast, and two hours later Erica likes Francesca's casts.
Reciprocity is the first sign that we are moving away from monologues to dialogues.
The two channels where people have most back-and-forth conversations are data and FarcastHer. The bar chart shows for each day the proportion of interactions that were reciprocal. For example, during the past 3 months there was 21 times 0 reciprocity in the data channel. We're gonna come back to the data-channel as it's interesting. < last sentence: leave in>
There is one caveat when making cross-channel comparison: The number of people participating in the conversation. During the 3 months there were 33520 interactions between 10386 casters in the nouns channel, compared to 4301 interactions between 1765 casters in the data channel.
Back to the bar chart (and not forgetting the aforementioned limitation): You can see that nouns (orange) and purple (green) peak at an earlier point than the other two channels. In both channels, the most frequent level of reciprocity is just above 0.1. Every 1 in 10 interactions in that channel is reciprocal. Everything else is a one-way flow. I cast, you like my cast, but I don't like your cast.
The distribution in FarcastHer is more evenly spread out, with a peak at 0.2. However, after this peak, the frequency doesn't drop drastically but slowly approach 0.
Where do we go from here? In social science network research, reciprocity is a stepping stone for everything. It's one of those variables you include in your prediction model because it always explains a bit about the predicted outcome. It's your foundation.
For casters this means: Do an effort and at least recast or like those who engage with your casts. If you just blindly recast and like everything, you're creating noise. It's like being at a concert at enthusiastically clapping and cheering at everything. /don-t do this. But step away from the monologues and start having dialogues on the channels. You'll benefit from it by getting to know people.
Result 3: Casting-circles aren't dense
It might be because Farcaster is now permissionless, but most people's casting-circles on Farcaster are very loose. With casting-circle I mean the people you normally interact with. These circles are loose as casters interact with a lot of different people. Our casting-circles are like revolving doors, with new names coming up over and over again.
I'm making this conclusion based on the low level of clustering coefficient, mapped on the y-axis in the chart below. Wait, what??? What's a clustering coefficient? Clustering coefficient measures if your friends talk with each other. If you are part of a 5-person friendship circle and the clustering coefficient is 1, then everyone talks with everyone. If your clustering coefficient is 0, then, well, there has been a huge blow-up and people aren't talking with each other.
A clustering coefficient of 1 in a social network as large as Farcaster is really difficult to achieve and not necessarily a good sign. It's like a typical Irish small village where everyone knows everyone. This has positive (people looking out for each other) and negative sides (conformity is valued above else).
Farcaster's low clustering of casting-circles is good news, as it means it isn't daunting for new-comers to join a channel. There's space for them to make a mark.
The high and lows in the trend probably is a sign of quality of casts. I'm assuming that high quality casts attract members to go more in depth and have a conversation around the topic. In essence, a move from a monologue to a group conversation.
In all channels, the average clustering coefficient increases towards the end and does not return to 0. This upward trend is most strongly visible in the nouns channel. While the magnitude of the clustering coefficient is higher for FarcastHer, at nouns the upward trend began earlier. Also, as nouns has nearly 5 times more casters than FarcastHer, a comparison based on the level of clustering coefficient isn't fair. Creating high density groups is easier in places with fewer people. Just think about everyone who lives in huge cities and laments the fact that they don't know their neighbors!
Result 4: There is change in the air
In all four channels there is an uptick in vibes. The level of closure exploded in Nouns and FarcastHer, but also purple and data saw an increase. Unfortunately, nouns kinda broke the chart with the number of vibing 030T triangles. Removing them didn't help much given the spike in FarcastHer at the start of February. The only solution was to create another metric: Closed vs open triangles
What the hell are closed and open triangles you might ask? Or if you are one of my kids you'll roll your eyes and glance at your phone. It's very simple: A closed triangle is good. It's vibes. It's you becoming friends with your friends. An open triangle is the opposite: No vibes.
By taking the ratio of open vs closed triangles it's easier to compare channels as we're looking at percentages. It's not a perfect measure, but as a trained scientist I'm always looking at limitations. Every limitation is another research opportunity, another grant 🎰.
The ratio Closed vs Open measures the tendency of vibes in the channel. It's about recognizing handles and pfp and adding to conversations in your friends participate in.
Looking at the ratios in the graph below, the first thing you notice is that the ratios are low for every channel. They normally hover around 0 and 0.2. But that doesn't really matter. What we are looking for is spikes and trends. Even better if they happen only in one channel. That is proof that the casters in the channel are vibing and it's not an ecosystem effect (I'm looking at you frames).
What do we got?
We might have an upward trend in all channels towards the end of the data sampling period (mid February 2024). As the effect is visible across channels, I'm assuming it's an ecosystem effect driven by frames.Timing works out. The assumption is that frames ingested energy in the ecosystem.
And we also got spikes. Oh so many spikes, especially in FarcastHer 💜
Data had the best day mid-January. It's a great channel. Part shilling queries and visualization, part Q&A and mentoring.
Nouns had a couple of spikes in December and January. It's a huge channel. Interactions can easily get diluted. What is missing for them is a better way to interact and surface conversations. But ccarella and seneca have a plan for that.
Purple had their best day towards the end of December. The graph looks like a constant low hum. No major spikes, but also no extended freezing spells at 0.
FarcastHer is only slightly larger than Purple, but shows a completely different pattern. While not topping the charts it has several spikes. Some small some tiny. It's like a roller coaster. But the good type.
I can't speak much about purple, nouns or even data. But I can tell you that at FarcastHer people spend time getting to know each other. FarcastHer meetups have helped to create a sense of "hey, I know you and what you are up to, and what your challenges are". In short, it looks like the norms we (aka all FarcastHers) set at the start and the effort that we put into building this digital third space is paying off.
What now?
Where do we, Farcasters, go from here? Before permissionless there was a huge discussion around qdau and wowow. These are signs that OGs wanted something else, a different third space, than any of the existing social media networks.
Curating spaces, be they digital or analogue takes intentions and efforts. Overly moderated channels create problems, making people anxious to get banned or unsure what to post. Lack of visibility on what casts get hidden and why do not help.
Channels current public nature invites an all-goes attitude. Everyone can post on every channel whatever they want. Couple this with other features (e.g., active badge) and experiments (degen) it can be hard to keep the forces balanced. How much to experiment without killing the vibe?
Gating channels can incentives more purposeful casting, but also replicate IRL social structures on Farcaster along socio-economic and cultural classes
Incentivizing recasting helps distributing your content, but can be seen as spam if done arbitrarily in other channels
I feel like I should leave you with some final words of wisdom. Like the practical implication section in every scientific article. Here it is:
Make friends on Farcaster. Cast for the long-game, not only for the short-term dopamine rush.
If you enjoyed the findings of this article, consider showing your support by collecting or sharing it.
If you made it all the way to the end, thank you 😘
Research used to be my bred-and-butter until ridiculous out-of-touch ivory-tower-like performance criteria killed the vibe. I'm talking researching talent and performance management while the university had no such processes in place. Not saying they are good, but the lack of walking the talk in some academic departments is mind boggling.
While writing this, it came to me how much I miss research, digging into patterns. I'll sit with the idea for another month. But the prospect of doing more netnographic research on Farcaster, interviewing channel leads, predicting casting behavior and channel growths and decline, excites me.
github file: https://gitlab.com/katerinabc/farcastersocialgraph
Overall Stats
References
Eric Quintane et al. (2013). Short- and long-term stability in organizational networks: Temporal structures of project teams. Social Networks, http://dx.doi.org/10.1016/j.socnet.2013.07.001
