Designing Trustworthy AI Systems with Cultural Considerations in Mind
While the use of artificial intelligence is becoming more ubiquitous across the globe, it is not equal in the ways that it impacts different demographics. As such, we believe it is critical for AI systems to be built (and regulated) with cultural considerations in mind.
In Mozilla Foundation’s recent research, we found “it’s not just the mindset of decision-makers that matters, but who is making those decisions matters. Tech has made strides in recent years to bring in new and diverse voices into product development, but we are still far from where we need to be.” This is one of the reasons why we support Fellows across the globe who are working on efforts to advance technology grounded in local needs.
Through the flagship Mozilla Fellowship program, we aim to empower people and bold ideas that can shape a more human-centered internet. We have fellows that are focused on mitigating a range of threats to internet health, and some of the Fellows this year have made great strides in raising awareness and proposing solutions to support the development of trustworthy AI systems that incorporate cultural sensitivities. Read more here:
Amelia Winger-Bearskin | CA, USA
While hosted by MIT Co-Creation Studio, Amelia sought to increase representation of indigenous contributions to thought leadership in the tech field. She also sought to create a model of ethical software dependencies that offers a pathway to inscribing community values and developer accountability into code.
Amelia’s impact: Amelia created Wampum.Codes, which is a framework for amplifying indigenous and native voices and values in the technology sector. Wampum.Codes includes a podcast, on which she interviews native and indigenous people who make cool things with new technologies. It also includes a project to develop a model of ethical software dependencies that offers a pathway to inscribing community values and developer accountability into code. Many organizations have signed up to be a part of wampum.codes workshops. There are also companies interested in embedding wampum.codes protocols into their weekly software development sprints. Launch here: Indigenous wisdom as a model for software development.
Coverage of Amelia’s work: Indigenous wisdom as a model for software development, Wampum.Codes at MIT’s Co-Creation Studio, Podcast, About the Wampum.Codes Podcast, Antecedent Technology, Wampum.codes SuperGroup Episode at MIT,ImagineNATIVE Official Selection Festival 2020 (Wampum.Codes podcast)
How you can take action with Amelia: Invite your team to a workshop by Wampum.Codes to learn how to include the Wampum.Codes value coding process into your software development sprint. More info here: http://www.wampum.codes
What Amelia is doing next: Wampum.Codes will continue as a project in residency at Stanford University, Amelia was awarded the Stanford Visiting Artist Fund in Honor of Roberta Bowman Denning.
Finally, what change Amelia has seen as a result of the fellowship: The development of the Wampum.Codes Award: Wampum.Codes next initiative is an award for native and indigenous high-school students to attend universities. The Wampum.Codes Award will create a pathway for native and indigenous students to attend postsecondary institutions, offering financial support alongside recognition and associated materials to elevate the participants profile in the job market. Wampum.Codes’ creation of an ethical framework for software development from an indigenous perspective can create a pipeline for scholarship and support indigenous people in computing. The Wampum.Codes Award is planned to launch for the 2021/2022 academic year.
Kostas Stathoulopoulos | London, U.K.
Kostas worked on addressing the following question (in his own words): “How open is scientific research when you cannot discover the most relevant outputs? My fellowship’s goal was to prototype academic search engines and exploration mechanisms that help researchers find work that completes their knowledge mosaic. [It’s important because] Timely and open access to novel outputs is key to scientific research. It allows scientists to reproduce, test, and build on one another’s work — and ultimately unlock progress.
Kostas’s impact, in his own words:
“I developed Orion, an open-source research measurement and knowledge discovery platform that enables users to monitor progress in the life sciences, visually explore the scientific landscape and search for relevant publications. I believe Orion’s web-interface will help users spot gaps and opportunities in research while the open-source code will assist science of science researchers with data collection and analysis tasks.” You can see more about Orion here: Orion’s open-source code, Orion’s documentation
Coverage of Kostas’s work:
- Orion: An open-source tool for the science of science
- Presentation at the Natural Language Processing and Data Mining for Scientific Text workshop
- Technical tutorial on Orion at the International Conference on Computational Social Science [part 1, part 2]
- Presentation at the Workshop on Open Citations and Open Scholarly Metadata
How you can take action with Kostas:
Do you want to get involved in Orion? Check out the contribution guidelines
What Kostas is doing next:
Kostas will be joining Util as a Natural Language Processing Engineer. I will be working on a system that leverages Sustainable Development Goals to evaluate how socially aware and environmentally friendly companies are.
Remy Muhire | Kigali, Rwanda
Remy's fellowship aims to address long-term internet health issues like digital inclusion through the use of open, emerging technology. Examples may include voice-enabled extension services to smallholder farmers to access digital financial products for underserved populations. Namely, those who are illiterate (currently up to 30% of adults in Sub-Saharan Africa). This is important because voice technologies are not a mere “convenience factor”. 30% of adults in Sub-Saharan Africa are illiterate. Currently, none of the main players in the global voice assistants market—Amazon’s Alexa, Apple’s Siri, or Google Home—support a single native African language. African languages lack enough data to train machine learning systems.
Remy’s impact, in his own words:
“So far we’ve collected over 1,000 hours of Kinyarwanda language voice datasets. Kinyarwanda is one of the fastest-growing languages on the Common Voice platform. The language has recently benefited from a speech recognition model which will open up unlimited opportunities to the local tech ecosystem and entrepreneurs. Part of one of the first use case to benefit from the model is “Mbaza”, an AI Covid-19 Chatbot which will provide information to the local community in Kinyarwanda.”
Coverage of Remy’s work:
- How Rwanda is Making Voice Tech More Open
- Deploy a React app to DigitalOcean using Github Actions and Docker
What Remy is doing next in his own words:
“After my fellowship, I will continue to work on my startup. Pindo a Cloud Communication platform for businesses. Pindo was established in 2019 with a strong belief in improving corporate experience in communication with their customers, by leveraging the power of Artificial Intelligence we want to build contextual assistants that really help customers with a focus on underserved languages.”
Petra Molnar | Toronto, Canada
While hosted at EDRi, Petra sought to address “how new technologies are used for border enforcement and immigration. From drones to AI-lie detectors at the border to big data predicting population movements, technological experiments are tested out of marginalized communities without governance and regulation, disregarding the real harms this technology creates.” This is important because “real people are impacted by tech experiments in migration.” You can see her final project here: Fellow Research: AI Systems at Borders Threaten Human Rights
Petra’s impact, in her own words:
“In many ways, work in this space is just beginning. We published our “Technological Testing Grounds” report with EDRi supplementing a new UN report on racism, technology, and borders that I also contributed to. We also launched a new collective called the Migration and Technology Monitor.”
Coverage of Petra’s work:
- UN warns of impact of smart borders on refugees: ‘Data collection isn't apolitical’
- Surveillance Won't Stop the Coronavirus
- Surveillance on the seas: Europe’s new Migration Pact
- Dispatch from a refugee camp during the COVID-19 pandemic
- Technology is the new border enforcer, and it discriminates
- Technological Testing Grounds: Border tech is experimenting with people’s lives
- Fellow Research: AI Systems at Borders Threaten Human Rights
- Project launch event: Discrimination at the Border: Launch of new UN report and the Migration and Technology Monitor
How you can take action with Petra:
“Migration and border technologies are on the rise and so are their adverse impacts. We need to hold states and the private sector accountable and abolish these harmful technologies. We also need more funding to support projects in this area.”
What Petra is doing next:
“I will be coordinating the Migration and Technology Monitor, a collective of civil society groups, journalists, lawyers, and filmmakers. We will monitor surveillance technologies and the use of AI to screen, track, and make decisions about people crossing borders, highlighting the far reaching impacts on people’s rights and lives.”
Finally, what change Petra has seen as a result of the fellowship?
“People are paying more attention to how AI intended to ‘manage’ migration makes it more arbitrary, unjust, and discriminatory.”
Josh Meyer | CA, USA
Josh is collaborating with organizations in East Africa to release open speech data and technology for African languages. As Josh notes, "speech technology is currently unavailable for most African languages. In collaboration with teams in Rwanda and Uganda, I am developing open speech resources for two East African languages: Kinyarwanda and Luganda.”
This work is important because “when speech technology becomes accessible in a new language, immeasurable value is created in that community.”
Josh’s impact, in his own words:
“I trained the first ever open Speech-to-Text system for Kinyarwanda, which is released to the community [https://github.com/Digital-Umuganda/Deepspeech-Kinyarwanda]. In collaboration with Digital Umuganda and the Rwandan government, this model will be deployed in a voice-activated chatbot to provide coronavirus information to Rwandan citizens. We also launched a data campaign for the Luganda language, spearheaded by AI researchers from Makerere University.”
Coverage of Josh’s work:
- Help create Common Voice’s first target segment
- Harvard Lecture: Applied Machine Learning for Embedded IoT Devices
- Kinyarwanda Speech data collection campaign
- Kinyarwanda trained DeepSpeech Speech-to-Text model
- Luganda Speech data collection campaign
- Luganda Data Processing Code
How you can take action with Josh:
Check out the Common Voice website, look for your language(s), and start contributing! If your language isn't yet on Common Voice, check out the Community Playbook for how to get your language included.
What Josh is doing next:
Josh says “My dream is to make speech technology available to everyone in their native language. I also hope to continue collaborating with Mozilla to democratize speech technologies.”
Finally, what change Josh has seen as a result of the fellowship?
He says “I have seen how powerful community action can be when harnessed for a common goal. A strong community can make the impossible possible.”