7/ There is a remarkable visual orientation in momentary symbols to grab attention in the digital realm 🎇

For example, the mask ban imposed in Hong Kong turned masks into a visual symbol that protesters and audiences shared with the international audience on Twitter to gain sympathy.

6/ Momentary symbols are volatile and can both spur and vanish in a matter of days as new events unfold💥

For example, lasers were practical tools to counter surveillance from authorities, until they became categorized as "weapons" after which they were used to visualize collective presence.

5/ Symbol construction is temporally contingent on concurrent evens (e.g., bans and arrests) ⏳

The figure shows spikes in mentions of the three selected objects over the span of the protest period - which often coincide with on-the-ground events in Hong Kong.

4/ We find that the Anti-ELAB 2019 protests highlight the empirical relevance of momentary symbols: objects with practical relevance experienced sudden spikes in symbolic-visual translations, but quickly returned to become practical tools once new symbols surfaced. Thus, we find that:

3/ Using a quali-quant inspired design, we construct actor-network inspired “translation snapshots” based on 1.6M tweets and object detection in ~670k images from the 2019 Anti-ELAB protests in Hong Kong to trace how umbrellas, lasers, and masks oscillate between symbolic and practical roles.

2/ We introduce the notion of "momentary symbols" to capture how objects briefly become dominant symbols for a collective, but just as quickly fade again. This stands in contrast to how the literature usually perceives symbols in collectives: as enduring visual elements over time.

I’m defending my PhD on Sept 9 at @sdu.dk 🎓

The defense is open to the public (in Odense from 10-12, followed by a reception).
If you want to join, please sign up here:
www.tilmeld.dk/phddefencere...

Excited to share the last 3.5 years of work - hope yo see you there!

@ddc-sdu.bsky.social

PhD Thesis submitted!
"Complex Dynamics in Digital News Ecosystems: Collaboration and Imitation in the Age of AI"

Grateful. Relieved. Maybe even a little bit proud.

#complexity #AI #news #platforms
@ddc-sdu.bsky.social

We also find the AI-generated articles are distinct from the human-written ones – yet are rather similar to each other.

We therefore envision LLMs as “gravity wells" that can, dependening on the initial structure of information, both compress and disperse articles in similarity space.

Furthermore, the effect of AI-driven imitation on news diversity is influenced by both the imitation strategy the agents use and their prevalence in the world.

We find:
• When content is originally homogeneous, AI-driven imitation increases diversity.

• But in heterogeneous settings, it might reduce diversity.

So, the impact of these "AI imitators" really depends on the baseline structure of the information environment in which they operate.

We combine 104K+ articles from Denmark’s digital news (2022) with large-scale scenario simulations to create “alternative realities” - or "multi-world simulations". We vary the prevalence of AI agents, imitation strategies, information heterogeneity, and prompting techniques within each world.

🚨New preprint alert🚨
"Autonomous AI imitators increase diversity in homogeneous information ecosystems" (with Oliver Baumann).

We study how AI agents that imitate human content could reshape news diversity in the immediate future.

A 🧵 on findings!

#AI #News #complexity #CSS @ddc-sdu.bsky.social

I am starting up a new project these days, where I really get to harness the power of the #OpenAI API. To this day, I am still mind-boggled by how much quality processing can be done for so little $$$.