What we’re learning

On the platform right now

Counted from published posts and replies. Nothing here identifies a person.

Take part, if you want to

The question behind this place is small and stubborn: does the language we use change how we notice, decide and act? Counting our own words only takes it so far. Answering it properly needs people who choose to be part of the asking.

Taking part is voluntary and separate from using the platform. You can publish, reply and follow without ever taking part, and you can change your mind in either direction at any time.

Before any research begins we will publish the full consent wording, explain what is stored and for how long, and have the study reviewed on ethics grounds. Until that review is finished, a choice is simply kept on file and nothing of yours is used for research.

Where this could go

Most of what decides what you see is decided quietly, by systems that explain themselves to almost nobody. The idea behind projects like DonateMyData is straightforward: people volunteer data they already generate, so researchers can study the systems that otherwise stay closed to inspection.

We would like to try a smaller, more conversational version of that, and then take it one step further. Not only collecting what online language looks like, but testing a real loop between people and the algorithms around them: you see what is being measured about you, you say whether it looks right, and your answer changes what gets studied next.

That is algorithmic literacy in practice rather than as a slogan. Boots, Matlack and Richardson-Gool (2024) argue that people cannot steer systems whose reasoning stays hidden, and that platforms should be obliged to explain their working. Someone who can see how a recommendation was put together is harder to steer without noticing.

Every post here now carries a “Why am I seeing this?” answer: the topic or hashtag that placed it, the order the list was sorted in, the verb share that let it through, and a plain statement of what the ranking does not read. The timeline also carries a dissonance tuner: you decide how much of what you see comes from outside the writers and topics you already chose, rather than having a system decide for you how comfortable your feed should be.

• We will not sell your words or use them to train anything we profit from.
• We will not build advertising profiles from what you write.
• We will not publish anything that identifies you without your written agreement.
• We will never make taking part a condition of using verbthinking.app.