Responsible and Societal AI Research

Avoiding the unintended consequences of technology through responsible design, development, and deployment. Addressing the production, and application of AI with a central focus on fairness, accountability, transparency, and the invisible labor embedded in AI infrastructures and its ethical implications within human society.

Technology has been a marker of societal progress since the invention of the wheel and even earlier. Technology is making our lives easier, more convenient, and it opens new opportunities for self- expression, communication, and economic progress. But technology can also have unintended consequences, often unforeseen even by the most well-meaning innovator.

Take a simple technology like the ‘infinite scroll’, which eliminates the need to keep pressing “next page” to reveal new content on a webpage. Originally proposed to help readers preserve their train of thought undisrupted by the inconvenience of a mouse click, it has since then spun out of control. On one side is the human who keeps scrolling to discover new interesting content, and on the other side is the content provider who strives to keep the content interesting. Together, it can spiral into addictive mindless scrolling and a steady narrowing of perspective – the rabbit-hole syndrome.

Or consider the ‘like’ button, which was created to bring people together and unite them in agreement. It too has given rise to unexpected and serious side effects, such as the diminished self-worth among those who crave likes that don’t materialize. Or take AI-based automated decision systems, which were hailed as a way to remove human bias from decision making, while at the same time also increasing efficiency, but eventually were found to be just as biased as humans, but now at scale.

Yet it is not only that the product of innovation can fuel addiction; innovation itself can also be addictive. In their thirst to win the technology race, innovators often lose sight of the adverse impact their striving might have on other seemingly unrelated issues, such as the enormous carbon footprints of large language models (e.g., GPT-3), reinforcement learning (e.g., AlphaGo), and block chain technology (Bitcoin).

There have always been adverse effects of technology. For example, some early studies showed that seat belts, when first introduced, increased the rate of accidents since they appeared to lower a driver’s perception of risk. One might even go so far as to say that the wheel itself should be blamed for the many car accidents it has facilitated. All of these are unintended consequences of technology.

More concerning is that the unintended consequences of technology are often unevenly distributed. Technologies that benefit some groups can simultaneously impose harm on others. This raises a critical question: who should bear responsibility and be held accountable for these unintended outcomes? For instance, the rapid iteration of hardware in data centers for AI-related computing has generated vast amounts of e-waste, contributing to environmental degradation and exposing communities in the Global South to serious health risks at dumping sites. Similar patterns can be observed in other domains, such as labor displacement, the erosion of human agency due to automation, and the amplification of biases within algorithmic systems. Responsibility in AI innovation, therefore, should not be limited to addressing consequences after they arise, but must be embedded throughout the entire lifecycle of technological development, from design and deployment to monitoring and governance.

It is clearly impossible and not at all advisable to stand in the way of emerging technology. Much better is to stay alert of its possible unintended consequences, study its possible impacts on society, and propose solutions and guardrails that can protect both society and innovators from these adverse effects. Sometimes these guardrails can be technology itself, such as frameworks for responsible, safe, and fair AI, or they can be rules, laws, and standards that provide humans with greater protection and control over how technology impacts them and how they can use it safely to improve their lives. The Department of Technology and Society is situated at this critical human-technology interface, committed to research, teaching, and practice that centers equity, human well-being, and human-centered design in the development and governance of AI.