Building on the community-driven due to the current lack of sophistication of machine learning systems to detect misinformation
Equal and easy access to fact-checking tool through chatbot for messengers enabling anyone to navigate the Infodemic around Covid-19
A color-coded credibility score based on maximum transparency clearly illustrates the decision of the Detectives at a glance and provides low-barrier support for users with lower level of media literacy
Community members are empowered in a multiple stage peer-review process and are incentivised to continue contributing through gamification elements
Adaptation of Open Source principles to share suitable datasets with scientists and/or journalists with the objective of shedding light on circulating misinformation in the "black box" of end-to-end encrypted messenger apps like WhatsApp.
Try our prototype here
Why do we need DetectiveCollective?
The WHO stated that “the 2019-nCoV outbreak and response has been accompanied by a massive ‘infodemic’ - an over-abundance of information – some accurate and some not – that makes it hard for people to find trustworthy sources and reliable guidance when they need it.” The virus is new for experts and society, which leads to substantial knowledge gaps and unknown variables during a time where our situation constantly changes. Out of fear, curiosity and a lack of sufficient knowledge, people can turn quickly to viral information on social media with sensationalist headlines and inaccurate information rather than consulting verified sources. Due to the algorithms of many social media platforms, viral messages - often dis- or misinformation - gain traction and engagement over truthful and validated information.
Misinformation does not stop at borders but clusters around common language, communities and similar levels of education. A diverse community could ensure that information is also fact-checked for languages less commonly monitored by fact-checking organizations and social media platforms. Originated from the German Hackathon WirVsVirus, DetectiveCollective is dedicated to help people navigate the mass of information all over Europe. Examples of misinformation regarding COVID-19 were fakes cures such as drinking bleach or accusations of Sweden having allegedly decided to sacrifice its old population while at the same time improving its pension funds for the case of collective immunity.
Fact-checking platforms such as Mimikama and Correctiv.org already verify information online. Unfortunately, these journalists from renowned or established media outlets are met increasingly with distrust by some strata of society.Their traditional newsroom structures can slow down the verification process - time during which misinformation can easily spread. Detective Collective can especially support people with low social media literacy, which are often most affected by misinformation and unknowingly distribute them further via social media and messenger apps. Conducting verifications through crowdsourcing - a community of people like you and me - might circumvent the issue of distrust against established media and politicians.
The European Union has proven with numerous initiatives that it recognizes the relevance of fighting misinformation. It is within this positive framework that DetectiveCollective wants to contribute to enabling users online.
How does DetectiveCollective work?
Users can easily forward messages to our Whatsapp chatbot - Detective Derrick. If DetectiveCollective has already evaluated the same content, Derrick communicates the result promptly. Otherwise, Derrick assigns the fact checking to the community.Considering GDPR guidelines, the submissions of fact checks will be anonymously.
The Credibility Score
DetectiveCollective is a community-driven platform where Detectives come together voluntarily to verify various content or “news items” The Detective’s evaluation about the credibility of an item is used to formulate and channel back a qualitative assessment (credibility score) to the user.
The evaluation process is firmly supported by the platform in order to help especially less experienced Detectives. Based on a repository of multiple questions, Detectives have to answer eight questions to fact check the news item (see below). Five essential questions (in bold) will always be asked whereas the remaining three will be sampled randomly from the repository.
The Detectives answer the questions in the repository by marking the respective news item with a color.
Red - not reliable
Red-Orange - rather unreliable
Orange-Green - rather reliable
Green - reliable
Hereby, we could possibly cooperate with existing initiatives and projects that classify websites of media outlets and blogs such as the Global Disinformation Index (GDI) and TrustedNews. The Detectives’ evaluation of the item will then be channeled back to the user with a message summarising the outcome and the color code the item received on average (see above).
To ensure that reliability of the results DetectiveCollective applies with a parallel peer-review process in which several Detectives work on a case at the same time and each evaluation is checked by a second independent and higher-ranking Detective.
Upon registration on the platform, the Detectives confirm to respect and adhere to a common code. This ensures the quality of the fact checking.
Continuation through gamification incentives
In order to ensure continued commitment on the platform, we create incentives through gamification to motivate members. This way, members gain points as “Detectives” for research that is carried out which allows them to rise in our level system and get access to new functionalities. In higher ranks, they can review the ratings of lower ranked Detectives or act as mediators in tricky cases.
Gamification offers the following advantages for DetectiveCollective:
Motivation - increase the fun factor
Transparency - rules and ranks are clearly visible to all interested parties
Togetherness - build trust within the community
Each Detective starts at level 1 out of a total of 10 levels. Upon successful completion of various tasks, a Detective receives points that lead to the next level. With increasing levels, there are new functionalities, more responsibility and prestige, as illustrated in the image below.
Points are awarded for:
Research and verification of submitted articles
Supporting other members in their research
Conducting peer reviews
Recruit new members
Master group challenges
Categorize and thematically assign messages
And much more
Fighting the spread of misinformation is a time-sensitive undertaking. For this reason, bonus points are awarded for particularly quick research for fact checking. However, the introduced verification process, including the peer reviews, ensures that, despite the high speed, the high quality of the research and thus the final evaluation is ensured.
Technical explanation of DetectiveCollective
To implement our solution we use the AWS cloud and its services. We use a postgresql database for storage and lambda functions for accessing the database and providing the required data to the APIs. For implementing the WhatsApp chatbot, we plan to deploy the Whatsapp Business API, while we use the Amazon API Gateway as the interface to the web application. We are also planning to leverage other AWS capabilities, e.g. Amazon Understand to develop a recommendation system to support the Detectives with verified reference articles.
Proof of concept architecture
As access to the WhatsApp Business API needs to be approved by Facebook first, we currently focus on developing an Angular.js web application that provides our functionality. Therefore, our current architecture looks simple but still provides the most basic functionalities to the users and the Detectives. Based on this core architecture, we will be able to iteratively extend our features, e.g. by connecting to the WhatsApp Business API, integrating user management using Amazon Cognito and handling asynchronous calls using Amazon SNS. If you want to have a look at our code, feel free to check out our GitHub page.
Challenges and lessons learned
Detective Collective originated from the German WirvsVirus Hackathon in the second half of March. During 48 h we formed a team of what were formerly strangers to each other, united by the idea to help people navigate information during the pandemic. We were overwhelmed by the responsiveness of people, willing to contribute their resources and develop the project from scratch. Within the team, we soon realized and appreciated each other's skills. Ever since, we have built on them to develop DetectiveCollective further. Being located in four different countries, it can be challenging to make space for the project aside from usual commitments but also provides opportunities to make use of the different time zones. The EU Hackathon gave DetectiveCollective another push to reach out to like minded individuals and it was no surprise to learn that while disinformation does not stop at borders, the commitment to fight those exists across borders exists as well. We joined forces with inspired individuals and groups to pool resources and make our concept and implementation even stronger.
What happens next
Detective Collective has the following potential to grow beyond its current horizon:
Development of additional functionalities (e.g. recommendation system to support the Detectives with verified reference articles) and integration of additional platforms (e.g. Facebook, Twitter etc.);
Growing community to cover more regions and languages;
Supporting science: DetectiveCollective is well suited to support dedicated researchers with examples and an archive of misinformation;
Post-Covid-19: DetectiveCollective is not limited to misinformation around Covid-19. The committed community can help tackle false information far beyond the pandemic;
Collaboration with private sector could provide non-monetary incentives for Detectives through goodies (e.g. free access to newspapers);
Out reach: Detective Collective is, in the framework of the EU Hackathon, reaching out to other like-minded projects. The ’Community-driven platform for identifying fake news’ and the ‘Whatsapp Fact-Check Channel’ want to mutually recognize benefit in exploring cooperation with DetectiveCollective, also beyond the Hackathon with the common objective to create a stronger product.
Try it out!
Check out our prototype here!
With the help of DetectiveCollective users have the possibility to evaluate news content anonymously and quickly - without prior registration or costs - according to their trustworthiness. The content is checked by an independent community of volunteers that is open to any interested citizen. DetectiveCollective builds on the collective interest to access reliable news content, a multistage verification process to identify misinformation and peer review by the community. In addition, our platform integrates the open source principle in the sharing of misinformation data, thus providing insights into false positives that circulate in the "black box" of end-to-end encrypted messenger apps like WhatsApp.