Part 2 of our 7-part series with the team at BlockScience features Senior Engineer David Sisson and Lead Social Scientist Kelsey Nabben.
BlockScience is a company that does analytical modeling in the distributed ledger (i.e., blockchain) field, working on complex cybernetic systems. David and Kelsie’s conversation explored a variety of topics, including:
- What is measurement?
- What can we measure?
- What do we do with what we measure?
- How can measurement data inform decisions?
David Sisson (DS) is a Senior Engineer at BlockScience. He was trained as a biologist and neuroscientist, and spent the first part of his career working in academia, but switched to technology about 25 years ago because “interesting problems engage me.”
Kelsie Nabben (KN) is the Lead Social Scientist at BlockScience, where she heads up the governance research team. Her focus is on the social dynamics of the engineering processes and challenges that the company navigates. She was trained as an ethnographer, and is presently completing her Ph.D. at RMIT University.
The interview was conducted by Eugene Leventhal, Executive Director of SCRF.
DS: For scientists, the bread and butter of any job is measurement. Scientists observe something and write down what they observe, and that’s effectively a measurement. From there they can make predictions, estimates, theories, or whatever else they want to do, but it’s always based upon measurements of the physical world. It doesn’t matter if it’s an old-fashioned thermometer that someone reads with their eyes, or a digital thermometer whose readings go directly into a computer.
KN: BlockScience works with complex systems, but they’re complex socio-technical systems, and the work is grounded in cybernetic principles.
DS: Officially, cybernetics is the science of control and communication. For BlockScience, this really means complex systems, systems of systems with some autonomy thrown in for good measure. Control loops are the best example, where something senses the environment around it and then interacts with that environment, all based upon measurements coming off a sensor. Ultimately, these processes are determining what kinds of actions an actuator should perform. It could be something as simple as a robot moving through a maze, or something as complicated as an organization of people, and how that organization is governed.
KN: Exactly. When the subject is “control of a system,” “control” in that sense is synonymous with governance, derived from the Greek word “kubernetes” which means “steering” (and is also the root of “cybernetics”). Moving from there to concepts that people are familiar with today, like “decentralization” and “decentralized autonomous organizations,” is based upon science’s ability to break them down into components that are dependably repeatable.
Now, social science takes a broader framing of all this. The idea of “second-order cybernetics” comes from early anthropologists who were there when this field was emerging [notably, Gregory Bateson and Margaret Mead]. They observed that the system designer was actually part of the system, and thus, they didn’t want to abstract the designer or engineer out of the system into a mere observer role. The designer and engineer are actually part of the system they’re steering, which means that they’re at a second remove, at the level of “a system of systems.”
This paradigm, also known as the “cybernetics of cybernetics,” goes beyond what is being measured in the system to ask why the system is producing these readings, what were the intentions of the people who made the system, what were the requirements they were designing for, and what were their assumptions? How do you make a “viable system” that is fully functional yet remains as autonomous as possible?
DS: The key tenet of biology is evolution. The survival of the fittest implies competition. Organizations don’t operate in a vacuum. They work alongside other organizations, with some of which they compete, with some of which they cooperate, and with some of which they do both things. So the organization needs to be “tuned” properly to be both effective and efficient. Communication is key to effectively managing semi- or fully-autonomous organizations. It allows independent operations teams, which understand their function better than anyone else, to self-govern within the larger whole.
KN: At the same time, there are clearly shared values and goals across the entire organization. That helps to create structure with the least hierarchy possible.
DS: From a data perspective, it’s important to take the frequency of what’s going on into account. Operations happen at a very fast pace. They answer the “how, when, and where” questions. Governance is slower, and is more about the “why and what.” You don’t want to change the “why” frequently because it confuses the participants about what the ultimate goals are. The “what” issues should only change as much as they need to, otherwise the organization is just churning. That separation based upon frequency of action is the main reason why the division of operations and governance is so important.
This is an important distinction, yet it may still be a bit too abstract for some viewers. Can you give a specific example from something in the real world that BlockScience has worked on?
KN: There’s a paper that Michael Zargham [co-founder of BlockScience] co-wrote on a DAO called 1Hive that he was involved with quite early in its existence. 1Hive adopted “conviction voting,” which is a mechanism for collective signaling of preference. That paper breaks down the core components in 1Hive and maps them against the VSM (value-stream mapping). We call that a “constitutional model.” One way to achieve a viable decentralized autonomous organization is by observing this pattern where the organization has a shared purpose that is usually written in a constitution,manifesto, or “terms and conditions” statement for the use of the platform.
1Hive has key pillars to its governance, including “conviction voting” or collective signaling of preference, which is how it passes proposals. What it’s governing is its treasury; its allocation of resources for funding open source software development within its own ecosystem. The working group that is granted the money is operationally autonomous. Ultimately, 1Hive is really measuring “what is value?” This is why when the question is asked, “are there any DAOs that actually work?” 1Hive always comes to mind.
DS: In the 1Hive example, “conviction voting” is nicely tied to the idea of governance being a low-frequency event. It’s specifically designed so that individuals stake tokens and the value of the stake increases over time. Nothing ever happens too quickly. As Zargham puts it, “there’s a built-in low-pass filter.”
Blockchains offer significantly larger amounts of data to work with than traditional closed systems. Are there any unique opportunities that arise from having so much data coming from open ledgers, or is this the same kind of endeavor regardless of the industry being looked at?
DS: Even in the open ledger world, the response is often the same: “We don’t want to share our data, so we’re going to encrypt it.” There’s no way to avoid that. There’s information that is going to be kept proprietary regardless. But what the ledger does allow is the ability to at least see the dynamics of all the interactions between addresses, allowing the creation of a temporal record of what went on. From there, it’s possible to build “estimates of state” at given points in time. In this data, there are measurements that can be turned into data, create models for predicting what’s going on, and make decisions based on that.
KN: Or evaluate what has happened as well. One interesting area is investigating the “goverance surface” of these systems–the boundaries that set the field of action where governance occurs. Researchers read the constitutions of projects, talk to people involved, and ask questions: What is your governance surface? How does governance work? What is possible? Who are the stakeholders? What actions can they take?
Then the data scientists can actually go look at that, look at the addresses and the actions that have occurred, look at the smart contracts and the parameterization of governance within those. That’s a fascinating breakdown of the applications of these methods in practice.
What are the research questions that you’re exploring based upon the knowledge that you are able to get?
DS: From a data perspective, there’s a lot of basic work that needs to be done in terms of classifying and categorizing entities based on their behaviors. It’s analogous to where biology was in the 17th and 18th centuries. Science needs good naming of things, a taxonomy or ontology, formal representations of the meaning that’s in the data. Keep in mind that there’s no magic in the data that comes off a distributed ledger. It’s bits; in and of itself, it means nothing. It’s the context that gives it meaning. Data itself isn’t sufficient for making decisions.
In the “meaning hierarchy” of data, information, knowledge, understanding, and wisdom (which is the same as “decisions”), gathering data as such tells you nothing. Hopefully, people reach understanding before they decide what they’re going to do next. For that, you need a schema. What are the units or numbers the data was collected in? What are the types of data that were involved? If you don’t know those things, the data is worthless. You need to create coherent views of the data in order to compare different sources intelligently.
KN: At BlockScience, people are navigating in real time the question, “what is computer-aided governance?” What is the role of computational mechanisms? That very broadly frames the research direction of BlockScience. BlockScience looks at governance from a variety of perspectives: pure computer science, engineering, anthropology and even humanities. How is that kind of collaboration maximized? How do we get more people from these very different disciplines united to improve governance?
DS: There’s a large body of knowledge that was developed in the web2 era around horizontally scaling systems with inefficient amounts of data. The web3 world was not set up well to allow for horizontal scaling. There was a lot of repetitive work that had to be done requiring bigger and faster machines. And yet, in the last year, there has been a lot of evidence of fruitful collaboration.
KN: Web3 has to be approached from an interdisciplinary perspective. Something is missing if researchers just talk to engineers, or just talk to social scientists. BlockScience as an organization has porous borders. Creating an interdisciplinary team automatically releases new energies.