The functions of scientific knowledge
In recent years, systems that track knowledge across the sciences, such as knowledge graphs that categorise knowledge concepts and AI-driven academic publishing search engines, have grown in popularity. These systems make knowledge mapping and search easier; they tell us about facts informed by evidence, hypotheses and results. However, these systems can't tell us which knowledge we could optimally apply to specific societal outcomes, i.e. the functions of scientific knowledge.
To understand and valorise the utility of science, we need systems to help us make structured claims and arguments about the possibilities of applied science. We want to be able to answer questions like:
What are the known constraints, both technical and commercial, on achieving societal outcomes?
What specific pieces of knowledge currently exist related to those constraints?
How might we uncover counterintuitive or non-obvious intervention points?
What kinds of expertise are likely to interact effectively with these pieces of knowledge, and who has this expertise?
How can we stimulate combinatorial innovation - getting people across domains to speak the same language and making knowledge easy to integrate?
To answer some of these questions, we have developed a new language that enables people to quickly get to first principles and communicate across domains in a rapid, programmatic way. In this article, we will describe the logic and implementation of our system, called the Outcomes Graph.
A protocol for applied science coordination
The Outcomes Graph is a collaborative, hyperlinked knowledge base that logs market and scientific research findings and points to the optimum path toward applying science to societal outcomes. We have implemented the Outcomes Graph in a collaborative version of the hypertext tool Roam Research using the Discourse Graph extension. The basic intuition behind the system is that we can write and synthesise knowledge as naturally as possible and then have the system recognise the important nodes and relationships.
Using the Outcomes Graph, we:
1. Characterise outcomes with precision and granularity
Unlike technology roadmaps or technology trees, the units of information in our protocol are Outcomes. Outcomes are positive statements of capability or achievement, usually suggesting a motivating factor we want to occur in the future. e.g. "to continue to live on earth without rising sea levels and turbulent weather systems wiping out civilisation".
Through a process we call scoping, we work backwards from the desired outcome to improve our understanding of the possible ways of achieving that outcome. We start by understanding the high-level state of the art, characterising the desired outcomes from the application of science, and who these outcomes will likely be important to (including customers, funds, users, beneficiaries, etc.). Of particular importance are four types of entities i.e. outcome nodes:
Constraints: An outcome which, today, we believe is not possible and for which our understanding is pretty complete, for example, something that blocks our pathway. e.g. Challenges, barriers, bottlenecks, thermodynamic limits.
Solutions: An outcome which, today, we believe is possible and we have high scoping completeness. e.g. Companies and/or startups that have a proven impact on the outcome we care about.
Hypothesised constraints: An outcome which, today, we believe is not possible but which we have a low understanding of. e.g. Failed attempts to apply a specific approach to an outcome, assumptions, or blindingly obvious beliefs about the status quo that govern the behaviour of incumbents, investors and entrepreneurs.
Hypothesised solutions: An outcome which today we believe is possible but which we have a low understanding of. e.g. Promising future directions that can become solutions or flipped hypothesised constraints etc.
Simply by typing, we then connect outcomes logically to and create relationships between them. For example, a potential solution may enable a constraint and companies are related to certain existing solutions.
Lastly, we set dynamic logical operators (AND/OR) between outcomes. This way, we can express how necessary and how sufficient certain outcomes are. For example, if there are many alternative Solutions (outcomes linked by OR connections), then this reduces Necessity, i.e the degree to which an outcome uniquely enables other outcomes. On the other hand, if there are many co-requirements (outcomes linked by AND connections), then this reduces Sufficiency, i.e. the degree to which an outcome enables other outcomes by itself. Outcomes that are more necessary and more sufficient are prioritised.
These contextual relationships are then accessible in every node, as demonstrated in the example of the critical material below.
2. Have evidenced discussions
Throughout the process, we want to know the state of the relevant pieces of knowledge that inform our arguments i.e. how ‘strong’ is this knowledge or its relation to the constraint?
We also want to know what data supports this particular set of statements, this specific framing, and how else that data could be framed i.e. what other pieces of knowledge do the data relate to?.
This is why we record evidence for every outcome we add to our system. Evidence can come from academic publications, industry reports, and interviews with experts and key opinion leaders. This allows us to argue about the tractability of certain approaches and decide if we need to scope further to increase our confidence.
3. Identify optimal paths to achieving high-impact ventures
Scoping happens at multiple levels. The logic, scoping language and its nested structure prevent us from going around in circles or getting stuck at one level. We can directly see the technology, market and value capture risks by evaluating the number of hypothesised constraints and constraints and also the opportunities via the number of hypothesised solutions and solutions.
We can also detect areas of neglect. For example, a group of outcomes may be more fertile to explore if fewer companies sit below them, or there may be some outcomes (hypothetical solutions or solutions) that are not connected to an existing company.
4. Discover opportunities for combinatorial innovation
Lastly, as we scope, we capture not only outcomes but also knowledge entities which effectively makes the Outcomes graph a Wikipedia for invention.
We can identify possible combinations of knowledge that have a high probability of achieving outcomes across completely unrelated knowledge silos and then specify the expertise required to combine this knowledge. In the algae example below, we can find linked outcomes and companies from 2 different DSV teams (agriculture & climate), four different projects (negative emissions technology and fertiliser), as well as ideas from our scientists and one of our portfolio companies, Aquature.
Toward collaborative venture science roadmaps
Attribution is critical to a well-functioning collaborative graph. The fine-grained tracking of individual contributions can allow an infinitely more granular taxonomy of individual and organisational expertise. This is a prerequisite for a more responsive, dynamic basis on which to allocate incentives based on individuals' expertise and contribution profile and the importance and probability of success of these specific endeavours. Eventually, this can lead to a more efficient and rapid assembly of consortia and a fairer way to allocate funding to proposed investigations, experiments and new ventures.
Harnessing the collective intelligence of venture scientists
Our community uses the Outcomes graph to describe the present state of the applied knowledge frontier, gauge critical pathways and bottlenecks, and find opportunities to move the frontier forward through venture creation.
We think Outcomes graphs (or parts thereof) could reduce the overhead to synthesis through reuse and repurposing over time, across projects, and potentially even across people. For example, imagine collaborators sharing outcomes graphs to speed up the efficient representation of arguments, claims and evidence to identify productive areas of venture building.
We would love to open up our system — bringing together philanthropy, government agencies, individual researchers, corporates and potential founders to work on a dynamic picture of the critical constraints, the paths of least resistance to curing cancer and neurodegeneration, removing carbon from the atmosphere and decarbonising hard-to-abate sectors, conserving nature and beyond.
We plan to build more features, including better search, automatic entity identification and linking, integrations with external databases, and onboarding of hundreds of venture scientists. If you are interested in using, developing and/or funding this further, please reach out to Eirini.