Our work

Open Terms Archive

Follow the changes to the Terms of Service

Services have terms that can change over time. Open Terms Archive enables users rights advocates, regulatory bodies and any interested citizen to follow the changes to these terms.

The tool is now operated by an independent collective.

Discover

Scripta Manent

Explore the contractual documents of the main online service providers and compare their evolution through time.

Scripta Manent is a first very simple online tool which enables you to compare changes between two dates of your choice. It's a valuable tool to empower consumers, lawyers, regulators (they are able to assess the loyalty of the platforms' commitments), lawmakers (they are able to analyze the rationale used by companies to circumvent their texts), journalists, law researchers, and even Legal Tech companies.

Try now

DisEncyclopedia

Improve your practices and toolset

An open and collaborative resource that documents best practices, tools and actors, allowing all of those who counter information manipulation to improve the competence of the entire ecosystem.

Explore

Information Manipulation Analyzer: search for information manipulation campaigns on Twitter.

Explore a narrative through volume, languages used, active users, associated keywords, media, coordinated inauthentic behaviour and an interaction graph.

See

Social Networks Graph Generator

Recreate the network of interactions of Twitter accounts on a given hashtag or word (or group of words).

See

Bot Finder: calculate the probability that a Twitter account is a bot

Defining what counts as a bot account is not as direct as it seems. For instance, some accounts are sometimes used by bots and sometimes by humans; these accounts are called cyborg and further complicate the definition. To help identify suspicious accounts, we have developed an algorithm that calculates the probability that a given account is a bot.

See

Open Terms Archive Experiments

Experiments are ongoing so as to produce use cases using Open Terms Archive data. To this end, two tools are currently under development : a script allowing to generate a dataset from an Open Terms Archive versions repository, and a python library easing the exploration process for these data.

See

Illegal Ads: Assess the legality of political ads

A crowdsourcing interface that allows citizens to view paid issue-based ads on Facebook and surface actors that do not respect the law. Currently only available for France, our code is open-source and we would be happy to collaborate with you to expand to other countries.

See

Collaboration chat: collaboratively detect, qualify and react to disinformation campaigns

A real-time chat that gathers actors, supports all the best practices identified in our encyclopedia, and offers the best tools as chatbots. If you are reading this, then we would probably like to have you on board.

Currently not available.

Bot Detection: identify clusters of bots on Twitter

A crawler that navigates Twitter follower relationships to identify clusters of suspicious accounts from a given seed account. It uses the free Twitter API and can thus be easily deployed.

Based on a manual annotation of over 400 accounts, we have also used that dataset to assess the reliability of usual bot detection tools such as Botometer.

See

Media-scale: compare an article's visibility on social media to other known events visibility in its local reference media to give some perspective to the number of reactions.

When fighting disinformation, analyses and fact-checks produced often relay - when they are not based on - figures to measure the impact of a content on a given society.

Among such data can be the number of reactions to said content, the number of comments it yields, or the number of times it has been shared.

However, this data us usually not contrasted nor compared. Rather than focusing on their quantitative aspect, we offer an alternative: to organize data according to a scale of relevance. Thus, in considering the topic of the content instead of how many times it was shared, one reaches a qualitative scale of reference.

See the source code