Research

Plenty of resources are being dedicated towards integrating artificial intelligence into our everyday lives, but far less attention is being paid to the tough but foundational questions. Our AI Harm Reduction research bucket has three main focus areas:

• Catastrophic Risk—how likely is AI to enable negative changes to human systems (such as, on the extreme end, causing human extinction)?
• Local Risk—how likely is AI to disrupt local communities through largely invisible but permanent changes, such as gradual changes to labor markets?
• Active Harm Reduction: How can we most efficiently mitigate these harms, with a particular eye towards educating and advocating for those who are most vulnerable to harm in an AI-enabled world?

Our approach is a combination of neutral, rigorous research, prototyping and building real-world tools, and engaging in targeted advocacy. We are currently designing a course on AI through the School of Engineering, and house an "AI and Labor" working group targeted towards key stakeholders from industry players to labor unions to individuals.

Social choice is the study of how groups should (and should not) make collective decisions. It involves topics such as simple elections, but also more complex deliberative decision making tasks, as well as behavioral, strategic, and group polarization aspects of collective decision making. Social choice is relevant to decentralized blockchain-based systems in two important respects: first, in terms of the governance mechanisms of the decentralized system; and second, in terms of the actual consensus algorithm that leads to block creation. 

Our primary research focuses on refining an abstracted blockchain model into a class of mechanisms that provide different trade-offs in efficiency (i.e., whether the most important transactions get included in a block), revenue (for block proposers), and fairness (to both transactions and proposers). Additionally, we study governance mechanisms with native dispute resolution.

Human health is a broad and fascinating area of study, ranging from concerns about day-to-day comfort all the way to the bleeding edge of longevity research. Unfortunately, a major barrier in this field is the lack of access to large, diverse datasets that capture the complex interactions between—to name just a few factors—genetics, environment, socioeconomic status, and lifestyle over time. Without access to this data, researchers struggle to make the key connections required to unlock the secrets behind healthy aging.

Our Health Engineering group focuses on democratizing access to this information, through technological or social engineering. We build tools to better interact with public health datasets (such as the NIH’s All Of Us database) and leverage citizen science to fill in the gaps that modern medicine doesn’t currently have the bandwidth to fill.

Progress is powered by incentives; the reality is that some projects lack built-in incentives. Non-monetizable public goods, for example, are low incentive, as are social initiatives that do not end in a product or publication. Often, the time required to construct even lightweight versions of these goods or services is outweighed by the risk of failure or lack of profit margin. As a result, the world is missing important pieces of infrastructure. 

Starting from first principles, our Social Redesign group reimagines these critical but overlooked human systems and builds minimum viable product (MVP) versions. Our hope is that by undertaking the early exploratory research and proof-of-concept modeling, other groups will face lesser risks and be able to pick up where we leave off

Theology and ethics are often artificially framed as counterproductive in the context of scientific advancement, meaning that these two fields rarely work side-by-side in any meaningful way. OpenLab directly rejects this assumption in favor of a framework in which theology and ethics influence and inform science and engineering, and science and engineering likewise empower theology and ethics. What does an engineering approach to spirituality look like? Possibly more interesting, what does a faith-based approach to engineering look like?

Our Ethics and Theology group is interested in advancing metaphysical research by taking inspiration from other disciplines. The traditional department-based university structure has kept these two fields apart for hundreds of years; our hope is to permanently change the way both operate in the world.

One of the most insidious threats facing US citizens today is the slow, nearly imperceptible creep towards authoritarianism. If history is any indication, we ought to be extremely wary of technologies deployed under the banner of public safety or general convenience. OpenLab is invested in considering emerging technologies and practices—from facial recognition and automatic license plate readers to the larger economy of surveillance capitalism—and asking: Does the worst-case scenario cost more than immediate safety and convenience is worth? 

Our Resistance Culture group explores cultural artifacts—fashion, art, music, and the like—as tools for adding friction into the surveillance economy. Our ultimate goal is to leverage the built-in plausible deniability of cultural engagement as a mechanism for slowing a potential descent into tyranny.

As DAOs and other decentralized coordination structures mature, new governance mechanisms become increasingly important to incentivize consensus “good” behavior, disincentivize consensus “bad” behavior, and reduce negative externalities. Governing a decentralized protocol—or, alternatively, using a decentralized protocol to govern groups of individuals—is a complex process with the potential to rewrite the rules of organizational management.

Our DAO Governance group draws on economics, law, computer science, and business to engage in both theoretical research and practical pilots. Our ultimate goal is to conceptualize and prototype new governance mechanisms.

Blockchains are a broad class of decentralized technology that have significant transformative potential. While most of the mainstream use cases have centered around digital finance, the concrete applications of blockchains in the real world is relatively under-explored. OpenLab believes that the promise of blockchains and decentralized systems goes far beyond investing and trading.

Our Applied Blockchains group examines how, for example, hardware or blockchain-adjacent software (cryptography, DAOs, etc.) might push decentralization further into public life. We experiment with injecting blockchains into real-world settings, as well as prototype new methods for cryptocurrency and DAOs to interface with the general public. Underpinning this research is a deep concern with the rising threat of authoritarianism—we believe that decentralized technologies offer one of the more promising solutions for avoiding a dystopian future.

Blockchains have a lot of great properties (decentralization, innovation, etc.) but can come with other headaches in certain contexts. Governments, for example, don't love their reputation for enabling criminal activity (although, it's worth noting that most studies support the idea that cash is still the preferred currency of criminals, and pointing out that most cryptocurrencies are not well-suited for illicit activities).

As an academic exercise, we are exploring what a digital currency system (more specifically, a smart contract execution engine) could look like if it were stripped down to the bare essentials. As it turns out, by doing things like removing most of the consensus overhead and building on faster storage, a smart contract platform can get quite fast and scalable. We are still architecting the actual system, but it our first draft can be toyed around with here: contractengine.stanfordprototypes.com

It seems inevitable that agents will someday have a hand (possibly a large hand) in consumer product searches. If this happens, one significant worry is that an agent may be captured by unexpected incentives that are not necessarily aligned with the well-being of the consumer. For example, it's highly conceivable that websites will start to offer referral fees to agents, or that middlemen will train their own agents around financial incentives. An agent may decide that recommending the consumer a product from a vendor for a flat $10 is less preferable than recommending the exact same product from another vendor that costs $10.05... but offers the agent a $1 referral bonus.

To combat this, we've prototyped an agentic search designed to keep agents aligned with consumer welfare. Instead of one agent recommending a product to a user (with no real checks or balances), two agents both search for products, and then present their arguments to a third agent, who will interrogate the choices and select the best one for the consumer.

Our primary codebase is still in progress, but a GUI version of this can be tested here: agenticproductsearch.stanfordprototypes.com

The US political system is (arguably, but plausibly) broken. There are many reasons for this, but at least some of them seem to have roots in our two-party, first-past-the-post system. This incentivizes individual voters to pick one of two main parties, and disincentivizes anyone who wishes to vote for a third party. The only way for a third party to be truly viable is to fix a massive collective action problem. Right now, third parties approach things backwards: They pick a platform (often for ideological reasons), and goes looking for voters who agree with it. That approach selects for ideological purity and against size and impact. It is a good way to build a group of committed people but a bad way to build an influential coalition that sidesteps coordination issues.

We don't know how to fix this, but perhaps taking a very staged approach is worth exploring. This might take the form of gathering tons of feedback on what voters actually want and care about, and then progressively asking people to increasingly signal a willingness to commit to the platform. Currently, we are in Phase 1: gathering information, which itself is valuable. Please feel free to explore the platform and, if you have the time and inclination, offer your thoughts on some of the positions. The prototype can be found here: thirdparty.stanfordprototypes.com

Two students from Harvard's Graduate School of Education asked how technology can help with human flourishing, eventually conceptualizing the Inner Animal game to increase self-esteem, cooperation, and self-actualization in school-aged children. The premise is almost deceptively simple: ask students to play an online game, together. Every time they go through a flourishing exercise, they will earn a seed that can be planted in a community forest. As the trees grow, students can water their trees or others' trees, eventually watching the forest become a (digitally) lush landscape as more and more exercises are completed. 

This project is ongoing, but a first prototype, built off the original ideas of the two Harvard students (who may opt into being named, if they wish!), can be found and tested at: inneranimal.stanfordprototypes.com

This prototype is intended for small-to-medium sized classes, and any teachers who wish to discuss testing it out in their classrooms are encouraged to reach out.