Who follows me, but I’m not following back?

I start by looking at who follows me, and whom I already follow back:

library(atrrr)
library(tidyverse)
my_followers <- get_followers("jbgruber.bsky.social", limit = Inf) |> 
  # remove columns containing more complex data
  select(-ends_with("_data"))
my_follows <- get_follows("jbgruber.bsky.social", limit = Inf) |> 
  select(-ends_with("_data"))
not_yet_follows <- my_followers |> 
  filter(!actor_handle %in% my_follows$actor_handle)

Now not_yet_follows contains 372 people! More than I thought. My assumption is that they are interested in similar topics and it would probably enrich my feed if I followed a chunk of them back. But how to decide? I came up with three criteria:

1. who is already followed by a large chunk of my follows
2. who has #commsky, #polsky or #rstats in their description
3. who has a big account, which I defined at the moment as 1,000 followers+

Number 1 and 3 are made under the assumption that popular accounts are popular for a reason and I’m relying on the wisdom of the crowd.

Who is followed by the people I follow?

To answer this, we need to get quite a bit of data. Specifically, I loop through all accounts that I follow and get the follows from them:

follows_of_follows <- my_follows |>
  pull(actor_handle) |> 
  # iterate over follows getting their follows
  map(function(handle) {
    get_follows(handle, limit = Inf, verbose = FALSE) |>
      mutate(from = handle)
  }, .progress = interactive()) |> 
  bind_rows() |> 
  # not sure what this means
  filter(actor_handle != "handle.invalid")

This data is huge, with over 450,000 accounts. So who in the not_yet_follows list shows up there most often?

follows_of_follows_count <- follows_of_follows |> 
  count(actor_handle, name = "n_following", sort = TRUE)
follows_of_follows_count

## # A tibble: 160,440 × 2
##    actor_handle              n_following
##    <chr>                           <int>
##  1 jbgruber.bsky.social              400
##  2 claesdevreese.bsky.social         352
##  3 rossdahlke.bsky.social            292
##  4 alessandronai.bsky.social         285
##  5 favstats.eu                       263
##  6 feloe.bsky.social                 263
##  7 jamoeberl.bsky.social             246
##  8 brendannyhan.bsky.social          226
##  9 fgilardi.bsky.social              225
## 10 dfreelon.bsky.social              224
## # ℹ 160,430 more rows

Unsurprisingly, I’m on top of this very specific list since this is a network around my own account. But let’s see who among my not_yet_follows list is popular here:

popular_among_follows <- not_yet_follows |> 
  left_join(follows_of_follows_count, by = "actor_handle") |> 
  filter(n_following > 30)

I put the people who have more than 30 n_following here, which is an arbitry number I picked, and ended up with 76 people I should look into.

Who matches my interest in their description?

Specifically, I look for a couple of key hashtags: #commsky, #polsky or #rstats in their description. These are the words I look for when checking out someone’s bio and it is very likely I want to follow them then. Looking for the keywords is pretty simple, since we already have the data:

probably_interesting_content <- not_yet_follows |> 
  filter(!is.na(actor_description)) |> 
  filter(str_detect(actor_description, regex("#commsky|#polsky|#rstats", 
                                             ignore_case = TRUE)))

Only 20 accounts fit this filter. Maybe I could find better keywords? But this is just a demo of what you could do, so let’s move on.

Who are the big accounts trying to connect?

We can look up the user info to see how many followers they have.¹

popular_not_yet_follows <- not_yet_follows |> 
  mutate(followers_count = get_user_info(actor_handle)$followers_count) |> 
  filter(followers_count > 1000)

Again the 1,000 follower number is arbitrary, but when I look at an account and see four figure follower counts, I still think it’s a lot. This gave me 80 accounts.

So what could I do now? Two ways to approach it:

1. let’s just follow them all if they fit these criteria:

lets_follow <- bind_rows(
  popular_among_follows,
  probably_interesting_content,
  popular_not_yet_follows
) |> 
  distinct(actor_handle) |> 
  pull(actor_handle)

follow(lets_follow)

2. More realistically though, I still want to have a look at the 136 accounts before following them.

This can be done relatively conveniently by opening the user profiles in my browser. I can do that with:

walk(
  paste0("https://bsky.app/profile/", lets_follow),
  browseURL
)

popular_among_follows2 <-  follows_of_follows_count |> 
  filter(!actor_handle %in% my_follows$actor_handle) |> 
  filter(n_following > 30)