The Bot Fighting Disinformation: The Story of Cofacts 

Written by Billion Lee

Image Credit: Photo by Jack Chen CC BY 4.0

Cofacts, an online fact-checking bot, has been trying to solve the “infodemic” for several years; and they are still reforming public trust in many ways. You can send suspicious pictures to the Cofacts chatbot and wait for Cofacts to fact-check the multimedia or photos for you. Fact-checkers devote themselves to news media and truthful information as best they can. Cofacts is an open-source project, and it is eager to have more new friends from global networks. As a multilingual chatbot, Cofacts speaks five languages: English, Japanese, Korean, Mandarin, and Thai. 

Why is fact-checking important? Now and then, in one of the numerous closed chat rooms composed of friends, colleagues and family members, lies like these may slip into the chatroom:

Homosexual people from Europe, US and Africa will marry Taiwanese gay people to reach Taiwanese free medical care for their terrible diseases

My country lobbied Pelosi to pay 94 million, about 3.14 million US dollars given to Pelosi by the Taiwanese government and its associated agencies: Visiting Taiwan =  A corrupted politician being bribed.

The first time we see stories like this, we might try ask google, collect enough information to debunk it, and reply in the chatroom with the information we have found in a way that does not hurt the feelings of chatroom members too much. But such hoaxes won’t occur just once. It may appear in another chatroom. It may change and be circulated again. Our fact-checking efforts in one chatroom go no further than the audience in that specific chatroom. They are buried in piles of chat history among numerous chat rooms as time passes. While fact-checking takes a lot of effort outside the chat app, sharing the hoax to another chat room takes a few button presses. We need a smarter way to face disinformation in closed messaging apps.

Unfortunately, compared to public platforms like Facebook or Twitter, hoax mitigation in closed chat apps like LINE, the most common in Taiwan, is sadly under-researched. We cannot put fact-check labels in end-to-end encrypted messaging apps. Network analysis of a message, which is effective in predicting if a message is a hoax, is not accessible in the context of closed messaging chat rooms. Nor can the general public analyse a user’s activity in closed chat apps to determine if the account is fake. Without these tools and effective methodology, the users in closed messaging apps are on their own when faced with disinformation.

Facing the problem on their own doesn’t mean that they need to fight the problem alone. Starting in 2016, Cofacts has tried to bring people faced with disinformation in messaging apps together via a message reporting chatbot and a crowd-sourced fact-checking community. Over 87,000 suspicious instant messages are reported via the chatbot and over 2,000 different contributors in the community. The community’s effort in Taiwan has been covered in mainstream domestic and international media outlets. Including PTS in Taiwan, Taiwan Panorama, NHK in Japan, Reuters, PBS, Al Jazeera, and so on.

 The chatbot

The most central thing Cofacts that tries to alleviate is the effort gap between forwarding a hoax and performing fact checks. Cofacts comes with a chatbot that sits right inside LINE messaging app, which has over 90% market share in Taiwan; anyone can freely add Cofacts chatbot to their contact list and forward messages. After the chatbot receives a message, it looks it up in the crowd-sourced hoax database to see if anyone reported similar suspicious messages previously. Then, it automatically sends the chatbot user the fact-checking reply for this specific hoax. It takes the same steps as forwarding the same hoax to a friend, and the difference is that it responds in seconds. Therefore, it is natural for users to look the message up in Cofacts before they share the same message with friends in the LINE messaging app.

After the user receives fact-checking replies from the chatbot, they can choose to forward the fact-checking response back to the chatroom they received the hoax. In this way, other people exposed to the same hoax can receive fact checks before they choose to forward the hoax to others. In addition, the Cofacts chatbot can also be invited to chatrooms to provide a friendly second opinion whenever a debunked non-political hoax enters the conversation.

The automatic fact-check response is composed on the Cofacts website, which publicly displays 87,000+ user-reported suspicious messages and welcomes anyone to submit a fact-checking reply. It feels like Wikipedia for counter-narratives of disinformation. Still, instead of everyone working on the same piece of writing, citizen fact-checkers on Cofacts can freely submit multiple fact-checking replies to one suspicious message. Fact-checkers may also review other fact-checker’s work so that the system can conduct compliant fact-checks steadily. Whenever a new fact-checking reply or a counter-narrative is submitted on the Cofacts website, the Cofacts chatbot can also send push notifications to those who previously reported such messages to update the audience with the information. The chatbot would display all fact-checking replies for the suspicious message all at once; the chatbot also asks if a reply is helpful to its readers.

Aside from traditionally marking trustful and false information, Cofacts allows fact-checkers to mark a reported message as “Contains personal opinion” and provide opinion pieces of different standpoints as reference data. This mechanism is unique to Cofacts (not seen in regular fact-checking organisations) because Cofacts believes only communication works for the public. In many cases, it is not only true and false but also regarding different feelings and opinions.

Cofacts relies on different walks of life and contribute the latest fact-check replies. Over 2,000 different contributors added their fact-check inside this fact-check platform, and over 87,000 suspicious articles were stored inside the Cofacts database. That is why it got the name “Cofacts”, Cofacts means Collaborative Facts (Fact indicates fact-checking). We hope to break the echo chamber rather than create new barriers during discussions. Cofacts do not have the authority to force you to trust that what someone said is correct; therefore, Cofacts would show you what other information is talking about. Cofacts wants to let everybody know: We are with you all the time.

What motivates the civic tech community? FUN!

Like most volunteer-oriented projects, Cofacts has difficulties with contributions and efforts conducted. Cofacts tries to make a fact-checking process an online game; you may watch a gif or receive a special bonus when you reach new goals; you can obtain a brand-new title when you combat a lot of fake news; you will see yourself level up when you accumulate enough EXP in the fact-checking platform. Gamification is the key to maintaining our platform and our system. Cofacts wants contributors to feel a sense of honour and respect for their contribution. You can review all the content you have generated, which is sharable. Johnson, Billion and other Cofacts experienced participants also host regular physical fact-check workshops to instruct new fact-check skills and chatting. By cultivating real friendship with goodwill, Cofacts hopes these events will last longer and bring joy to our volunteers.

The power of openness

Cofacts is an open-source software project providing everyone technical support, tool kits, database, and open API. This enabled third-party downstream chatbots to appear in Taiwan. Together these bots can send fact-checking replies on Cofacts to more than a million audiences. The Cofacts database provides statistics of what and when messages are being circulated, which is valuable for professional journalists, investigators, and amateur fact-checkers. The Cofacts database provides fact-based real-world data for journalists like Readr, and The Reporter, so they can focus on writing nicer reports and take advantage of this system to easily improve their workflow. Researchers like Doublethink lab, IORG, and IFTF also publish analyses and insight on Taiwan’s disinformation landscape, referring to data provided by Cofacts. Lastly, Cofacts’ source code has been replicated and forked to Cofact in Thailand. The experience and the technique development can be learned and duplicated in other organisations and countries. 

Disinformation affects everyone, but everyone can become part of the solution. This is a simple idea that powers Cofacts and many other g0v projects. Although Cofacts has had experts contribute and have worked with other organisations, the fact-checking process is open to all. Disinformation breeds distrust and polarisation, but collaborative fact-checking breeds trust and collaboration. When governments get too involved in fighting disinformation, it can look like an infringement on free speech. That is why it is so important for civil society groups to get involved. The process can be slow: disinformation spreads earlier than fact-checking, but just like the tortoise and the hare, our strength lies not in our speed but our innovation and resilience. Cofacts is fighting the long fight, and it is only possible by creating a structure that is open to anyone and for everyone.

Billion co-founded Cofacts in 2016, is a contributor of  g0v cvic tech community in Taiwan. Billion has organized the voluntary fact-check community and contributed most of the database’s content, providing media literacy and fact-check workshop classes in Taiwan.

This article was published as part of a special issue on “The g0v decade”

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s