Hope in Action: Open-Source Innovations for Information Integrity
Highlighting how open-source is helping to address mis- and disinformation around the world
The Digital Public Goods Alliance (DPGA) and the United Nations Development Programme (UNDP) are deeply concerned about the impact of information pollution, especially in fragile or conflict-prone contexts. That’s why, in February of this year, we launched a global call with the goal of discovering and highlighting innovative open-source solutions that can help promote information integrity. This initiative is in support of the Nobel Prize Summit, “Truth, Trust and Hope”. The call received an incredible response from individuals and organisations worldwide, with 99 technologists, innovators, and change-makers from diverse backgrounds and sectors submitting their open-source solutions and concepts. Today, we are excited to share the 9 solutions that have been selected based on their innovative approaches. Collectively, they are dedicated to fostering a more trustworthy digital experience, ensuring transparency and accountability, mitigating the role of disinformation in conflicts, and providing better access to reliable information.
Information pollution knows no borders. It affects societies worldwide, distorting facts, eroding trust, and damaging democracy and human rights. In countries with weak governance systems, or those in conflict or crisis, often in the Global South, inaccurate and misleading information can be weaponised to have even more devastating effects. However, the DPGA and UNDP have a firm belief in the transformative power of technology and aim to provide hope and accelerate action against mis- and disinformation by making relevant open-source solutions available to all.
Well-designed open-source tools can offer distinct advantages in addressing the challenges posed by information pollution. They allow transparency and collaboration while facilitating easy adoption and adaptation to meet the specific needs of different contexts. For this reason, we believe that these tools have the potential to be used or adapted and scaled relatively easily and quickly to have significant impact where needed.
Collaborating with an advisory panel of experts from academia, technology, civil society, public policy-making, and the private sector, we identified the most needed and unique approaches among the submissions. Those selected address information pollution through various approaches, including combating deepfakes, analysing online news media, verifying crowdsourced data, monitoring tech companies’ legal terms and their impact on users, improving access to government policies, and gaining insights into the influence of digital technologies on societal conflicts. By empowering users, these solutions contribute to combating disinformation and fostering a more informed and resilient information ecosystem.
Congratulations to the following selected solutions:
Ad Observatory
Incubated at NYU, Ad Observatory is a powerful tool that aims to enhance transparency in digital political advertising on Facebook during elections and provides valuable insights into ad topics and advertiser intent. This tool can play a crucial role in combating disinformation by shedding light on political advertising practices and fostering public awareness and accountability.
Ad Observatory was utilised by reporters from major US news outlets in 2020 and 2022, helping to inform readers about digital advertising in elections, uncovering spending patterns, and misleading advertisements. It has also helped explore the role of algorithms and advertising in society for researchers in Australia. Ad Observatory’s advanced language capabilities facilitate analysis in both Spanish and English. Moreover, they have made their back-end open source and are currently working on developing an open-source front-end toolkit, which is expected to be completed by the end of 2023. In 2024, Ad Observatory plans to provide the public with practical transparency about digital advertising in the US presidential election, and partner with European organisations to develop a European version to provide transparency for the European Parliamentary elections.
Deep Fake Fingerprint
The proliferation of deep fake media has led to an array of devastating consequences ranging from misinformation campaigns to false warfare propaganda, and fraudulent activity. In particular, the recent explosion in accessible voice cloning technology has meant that convincing fake audio content can be created at the click of a button by anyone with an internet connection.
Selected in the concept category, and led by researchers at UC Berkeley, the Deep Fake Fingerprint is being developed as a research methodology to learn and provide a unique identity fingerprint for any given person using only short snippets of media. Focusing on audio content, the goal of this work is to create a distinctive pipeline for individuals that can ultimately be used to authenticate their voice. The developed method ingests short snippets of audio as input and outputs a user-specific pipeline to analyse new audio clips for authenticity. This technology may be particularly valuable for world leaders and public figures that nefarious actors may be incentivised to target. Further, it is expected that this work can be more broadly applied to the general public and allow everyday internet users to protect themselves and rebuild trust in what they see on their screens. Moving forward, the team is aiming to submit their work to a leading information & forensics conference, before undergoing a rigorous testing phase to further explore the robustness of their results and prepare the technology for production.
Feluda
Feluda, developed by Tattle, addresses the challenge of combating mis- and disinformation in the context of social media platforms that heavily rely on images and short-videos, particularly in regional languages in countries like India. Traditional methods of analysing content through URLs are insufficient as they are often absent in photos or messages, making it challenging for researchers. Feluda helps make multimodal analysis possible by using visual and semantic similarity across modality. Feluda is an engine designed to understand native multimedia content in Indian languages. It can be used to conduct image and video searches on Indian fact-checking sites, generate analytical reports on content circulating on platforms like WhatsApp, and answer questions like what themes are present in a multimodal dataset. The engine is a valuable resource for any team seeking to understand large volumes of multilingual and multimodal content, aiding in the fight against information pollution particularly for fact-checking, research, and data science. Feluda’s impact is showcased in this case study exploring information chaos on WhatsApp during India’s second COVID-19 wave. Sustained funding in addition to collaborating with journalists and fact-checkers to help them find important stories, track narratives, and move beyond the study of text-only social media, can help enhance Feluda’s impact.
Media Cloud
Media Cloud empowers researchers, journalists, and organisations to combat disinformation by providing access to a vast repository of online news media content from around the world. This system offers free and open tools for research and analysis, enabling the study of digital news and online information flows on a global scale. By understanding and tracing the amplification of stories across online news media, it becomes possible to analyse and mitigate pollution within the information ecosystem. Moreover, Media Cloud facilitates the examination of reliable media sources and their narratives, contributing to the study of information systems. The platform’s user base consists of academics, journalists, independent researchers, and non-profit organisations and foundations. It has recently developed a search interface that allows users to explore content from various social media platforms, such as Twitter, Reddit, and YouTube, further enhancing research capabilities. Media Cloud is growing to accommodate the increasing volume of information, and creating easier to use interfaces to support non-media experts utilising its features. Media Cloud is also looking to grow the supported languages within the system, as well as increase the metadata that is retrievable about content and publishers.
Open Terms Archive
Open Terms Archive is a digital common empowering journalists, regulators, lawmakers, researchers, and activists to understand, respond, and influence the rules of online services. It publicly archives the terms and conditions of services in different languages and countries, making them easily readable and highlighting changes. This allows individuals to uncover unfair practices, ensure better protection against misinformation, hold big platforms accountable, and design effective regulations. Open Terms Archive connects and empowers individuals and organisations to collectively improve the transparency and fairness of online platforms to help foster a healthier digital experience.
After its focus on information pollution and consumer protection, Open Terms Archive is hoping to work with funders to leverage crowdsourcing to track the terms of generative AI services, an industry that is rapidly evolving, mostly unregulated, and that will massively impact many sectors.
Phoenix
Phoenix, by Build Up and datavaluepeople, is a solution designed to address the negative impact of digital media on societal divides and conflicts. It aims to provide peacebuilders with a deep understanding of the divisions present in digital media and their real-world consequences. Phoenix achieves this by scraping and organising digital content, utilising artificial intelligence to automate classifications created in collaboration with peacebuilders. This enables detailed analysis of digital conflicts and generates actionable insights that peacebuilders and mediators can use when facilitating decisions. By recognising the interconnectedness between online and offline events in conflict settings, Phoenix empowers peacebuilders to bridge the divide and foster constructive dialogue. Phoenix is iteratively improving based on peacebuilder and mediator experiences, and the team is currently working on a self-deployable tool and on building a community of open-source developers that can contribute to its sustainability as an affordable, non-commercial solution accessible to local peacebuilders around the world.
Querido Diario
Querido Diario, developed by Open Knowledge Brazil, addresses the challenge of accessing and analysing official decision-making acts throughout Brazil’s cities. With no centralised platform available, the only reliable source of information is in the closed and unstructured PDF files of official gazettes where they are published. To tackle this information gap, Querido Diario’s robots help collect, process, and openly share these acts. Launched over a year ago, it has grown into a comprehensive repository with more than 180,000 files, continuously updated with daily collections. Querido Diario helps combat information pollution by providing a transparent and reliable source of data that can be used to fact-check and counter false narratives, enabling informed analysis and promoting accountability. The primary users are researchers, journalists, scientists, and public policy makers and it helps benefit various sectors including environmental researchers and journalists, education NGOs, and scientists working with public data. Today, Querido Diario’s coverage reaches 67 cities, where 47 million people live. The next steps involve scaling up to include all 26 Brazilian states and at least 250 cities. The project aspires to incorporate Natural Language Processing models and integrate its data with other public datasets, helping users contextualise information even more.
RegretsReporter
RegretsReporter, a browser extension built by the Mozilla Foundation, powers the largest ongoing crowdsourced investigation into YouTube’s recommendation algorithm. With participation from over 70,000 individuals from 191 countries, RegretsReporter allows people to donate data about harmful videos that are recommended to them on YouTube. Mozilla’s analysis of this data has uncovered dangerous patterns in the algorithm’s behaviour, including the frequent promotion of content that violates the platform’s own Community Guidelines, a concerning issue that is more prevalent in non-English speaking regions.
YouTube’s recommendation algorithm is notoriously opaque. RegretsReporter disrupts this lack of transparency by harnessing community participation, shedding light on how the algorithm may contribute to information pollution. Furthermore, the project’s recent extension release introduces a unique feature that enables users to contest and challenge problematic recommendations, leading to a reduction in information pollution. RegretsReporter has powered research and investigations that have been cited in European regulations, in US Supreme Court cases, and in shareholder resolutions aimed at holding YouTube accountable. Now, the Mozilla team are releasing anonymised public datasets from RegretsReporter to enable people from around the world to conduct their own impactful investigations into YouTube’s algorithm.
Ushahidi
Ushahidi is a global not-for-profit technology company that gives citizens the tools to generate data on the ground – raising their voices, influencing change, and mobilising support. Ushahidi builds open-source software that helps people gather, analyse, and act on data, whether it’s about climate change, elections, or any other issue that affects them. Their flagship product is an integrated data crowdsourcing and mapping platform that plays a crucial role in combating disinformation by ensuring information integrity, particularly in critical situations like elections and humanitarian disasters. It utilises criteria, such as location and time, to validate crowdsourced data for accuracy. Collaborating with partners, Ushahidi verifies urgent information and facilitates prompt responses. The platform serves a diverse user base, including the general public, first responders, and international organisations, enabling them to collect and analyse information, report incidents, and gain situational awareness. Ushahidi’s tools empower users to address important issues, promote citizen participation, inform decisions, and drive systemic change.
Since its inception, the premise of Ushahidi’s work has been raising the voices of marginalised groups to ensure that their lived experiences influence the change they want and need to see in the world. Ushahidi aims to raise 20 million additional voices of marginalised groups by 2026, with the main goal of achieving meaningful, long lasting change through knowledge based on inclusive and truthful data.
We express our gratitude to our advisory panel of experts whose extensive experience and insights were invaluable in assessing the submissions. An extended thank you goes to Craig Newmark Philanthropies, the Government of Norway, and Omidyar Network for their support of this initiative.
We eagerly anticipate the opportunity to promote these solutions at the Nobel Prize Summit, a prestigious gathering co-hosted by the Nobel Foundation and the National Academy of Science in Washington, DC, from May 24-26, 2023. This invaluable platform will not only showcase these innovative solutions but also provide an opportunity for potential partners and funders to learn more about these invaluable tools and how they can support them.