Machine Learning & Society Topic Cards
The interaction between Machine Learning-based systems and their social context can be examined through the lens of different inter-related topics. In our work, we have found the following 8 topics to constitute a helpful way to organize resources:
Understanding: Openness and the Science of AI
As AI is becoming ever more ubiquitous, increasingly more builders and affected stakeholders need to understand how it works, what it can and cannot do, what trade-offs are involved in developing the technology, and how it can be leveraged or improved in particular contexts. This requires sufficient visibility and a thriving research ecosystem that is inclusive of perspectives outside of those of the developers working within the very best-resourced companies.
Making informed decisions about AI systems requires understanding how the technology works, what development choices are available to meet certain goals, and how they trade off different priorities. Approaching AI as a science means upholding scientific integrity, which includes reproducibility, verifiability, and increasing the breadth of people who can use the technology and contribute to scientific development.
Developing: Openness, Business, and Competition
“AI” and Machine Learning systems are already ubiquitous, deployed in social media and all kinds of digital services. AI can be compared to “Software 2.0” - AI literacy is becoming part of the basic set of skills to build any kind of technology. Relying exclusively on AI systems that are developed and served by a handful of companies would have strong negative impacts on innovation and competition.
Having access to data is a significant competitive bottleneck, including: data about how people are using AI systems - data about what AI systems are better or worse out - use data and feedback from users - proprietary data to fine-tune models - internal databases for RAG-like techniques. Sending all of that data to a few companies further centralizes their role in adapting new technology. In addition to enabling anti-competitive practices, this also limits the breadth of technology that can be developed - better to have smaller companies work on their own hundreds of thousands of use cases than to have central entities decide what is worth deploying and providing a “just OK” unique solution for everything.