Kate Starbird
Excited for the publications of our paper (w/ Kaitlyn Zhou and Emma Spiro): Spotlight Tweets: A Lens for Exploring Attention Dynamics within Online Sensemaking during Crisis Events. The paper, which looks at sensemaking during the 2018 Hawaii missile crisis, makes several contributions, including: presenting the concept of “spotlighting” and introducing our cumulative graph technique, which reveals how audience exposure affects the propagation of social media posts.
https://dl.acm.org/doi/10.1145/3577213
Spotlight Tweets: A Lens for Exploring Attention Dynamics within Online Sensemaking during Crisis Events | ACM Transactions on Social Computing

In this paper we introduce the concept of a spotlight social media post — a post that receives an unexpected burst of attention — and explore how such posts reveal salient aspects of online collective sensemaking and attention dynamics during a crisis ...

ACM Transactions on Social Computing
Jan 19, 2023 at 4:27amWeb
Show threadKate StarbirdJan 19, 2023
The paper explores how people made sense of the false missile alert in Hawaii in Jan 2018, introducing the concept of a “spotlight” social media post, where a post receives unexpected attention, often because a high-follower account bestows attention on a lower-follower account. We also show how spotlighted tweets differed from other tweets (including other highly-retweeted tweets from high-follower accounts) during the missile crisis.
Show threadKate StarbirdJan 19, 2023
Not surprisingly, spotlight tweets (tweets from typically low-follower accounts that received an atypical amount of attention for that account) were more likely (than other tweets) to be posted by someone who was in Hawaii and to contain personal accounts (including first person account) of receiving news about the alert. Popular tweets (highly retweeted tweets from large following account) were more likely to politicize the crisis.
Show threadKate StarbirdJan 19, 2023
The paper is also the original work that introduces our “cumulative graph” visualization technique for illuminating audience exposure in tweet propagation. We have been using this technique publicly for years (while the paper has been in revision), but the technique was originally developed by Kaitlyn Zhou, the first author. Here’s a recent example, showing how reports and rumors about voting issues in Maricopa County went viral in 2022. http://faculty.washington.edu/kstarbi/Cumulative-Maricopa-2022-going-viral.html
Bokeh Plot

Show threadKate StarbirdJan 19, 2023
This “cumulative graph” technique shows how a tweet or set of tweets about a topic spreads, highlighting the role of high-follower accounts. Time runs left to right. The y-axis shows total spread. Tweets are plotted and sized relative to follower count. Tweet shapes here show type (original vs. quote vs retweet). Tweets can be colored by author or content features. In this case we’ve colored tweets that retweet or mention two influential accounts. Here’s a view annotating high-follower accounts
Show threadJon (now at neuromatch.social)Jan 19, 2023
@katestarbird very interesting ... "spotlight tweets" are a very interesting concept, and the visualization is quite illuminating! (also excellent alt-text on the diagram). The four exemplars are very well-chosen too, an excellent paper!
Show threadKate StarbirdJan 19, 2023
@jdp23 Thanks Jon! It went through 5 revisions, so at this point, it's like 3 papers smashed together. I'm happy to hear you enjoyed it!
Show threadJon (now at neuromatch.social)Jan 19, 2023
@katestarbird it's such an interesting phenomenon the way a tweet by a "random person" can suddenly go viral, and looking at it in this context is so important. And of course seeing it here first, I thought about how the whole discussion in 7.2 applies both with the disinfo-optimized "Twitter 2.0" and in the post-Twitter fediverse / post.news / etc. environment. Timely!
Show threadUlrike HahnJan 19, 2023
@katestarbird the speed of information spread is terrifying..
Show threadamincoJan 19, 2023

@katestarbird

Brilliant! If only these cumulative graphs could track the spread in real-time...

BTW, as of 2023, does Twitter still allow access to researchers?

Show threadKate StarbirdJan 19, 2023
@aminco (1) We can usually put these together with about an hour lag time. In theory, platforms could build tools that would allow people to explore these trajectories for a piece of content or a topic. (2) For now, yes. But it looks like they are shutting down the V1 API and the V2 API has much reduced limits in terms of how much content we can collect.
Show threadamincoJan 19, 2023
@katestarbird
I wonder how that would look if displayed using log scale.
Show threadDoug BostromJan 29, 2023

@katestarbird

I'm guessing data for this work is gathered via Twitter's provisions for academic research.

Hence curious to know: is Twitter showing signs of going silent as a data source for work of this kind, given the general degradation of the system?

Important instrumentation.

Show threadKate StarbirdJan 30, 2023
@Doug_Bostrom Twitter has announced plans to power down (next month) the V1 API that we’ve been using for years. The V2 API (for which we have “academic” access) remains active for now, but it has limits that will make it difficult to do some what we have done in the past. We’re working to adjust our methods and infrastructure for now, but are also wary that academic V2 access could also end at any time.