Memes in academic literature are defined narrowly

Outside academia, most discussions use Dawkins' ambiguous 1976 definition of memes as "units of cultural transmission, or units of imitation"

It encompasses a broad range of cultural elements, including abstract ideas (like God), textual content (such as nursery rhymes and jokes), and behavioral practices (like Christian rituals).

However, academic research has largely adopted a different understanding of memes. Shifman (2013) defined Internet memes are not as individual ideas or formulas, but as groups of content items that were created with awareness of each other and share common characteristics

Notice how the focus shifted from viewing memes as individual pieces of content to understanding them as part of larger meme families.

Notice that new definition excludes ideas, or texts or even catch-phrases as there are no groups of varied but similar content for them.

As most academic literature uses Shifman definition, the term "memes" in it includes only pictures, gifs and videos spread on the internet through imitation. We'll use Shifman definition in this thread as well

How to study memes quantitatively

Families and Networks of Internet Memes: The Relationship Between Cohesiveness, Uniqueness, and Quiddity Concreteness

Shifman's definition of meme families creates an interesting paradox: memes within a family must be different enough to be distinct, yet similar enough to belong together. But to maintain this balance of difference and similarity, we need to somehow measure it, don't we?

It's easy to do so intuitively - after all we can trivially recognize from which meme-family (or families) is any given meme. The challenge is measuring this objectively using only general meme attributes, without relying on our intuition.

The paper suggests we can do it by encoding each meme as 39 generic attributes - from ideas and ideologies of the meme to the form (e.g. close-up shot) to tone and style of communication.

Then constructing a graph of memes where weight of the edge between two memes depends on how many generic attributes they share. Turns out that meme families form distinct clusters in such a graph, as we've expected

These findings suggest we can study memes objectively using quantitative methods - if we encode them properly. And with new Vision Language Models (VLMs), we might even automate this encoding process!

Ups and downs of disinformation research

Disinformation and Echo Chambers: How Disinformation Circulates on Social Media Through Identity-Driven Controversies