Not-knowing about actions and outcomes

22/6/2023 ☼ not-knowing

Warning: Ideas and analysis in this post are still being baked. This is the reading for the 22/6/2023 discussion that’s part of my monthly Interintellect series on not-knowing. The next discussion will be on 20/7/2023 and will focus on not-knowing about causation.

Actions are physical or intellectual things agents can do, and outcomes are the results of actions agents take. Much complexity lies within this apparently simple pair of definitions.

Both actions and outcomes have properties. These five properties seem most relevant, and each is a continuum

1. Size: big ←→ small
2. Specificity: specific ←→ general
3. Independence: isolated ←→ enchained
4. Overtness/visibility: explicit/apparent ←→ implicit/invisible
5. Comprehendedness: well-understood ←→ not understood

To think clearly about actions and outcomes requires this vocabulary for describing their properties. In turn, thinking about not-knowing of actions or outcomes is much easier when we can think and refer clearly to them.

Properties of actions and outcomes.

Actions

Actions can range from the very small and trivial (“let go of the banana“) to the large and consequential (“fund an expedition of explorer ships to discover sea routes to Asia“ or “invest in R&D.”)

To illustrate some of the properties of actions, consider this set of three linked examples:

1. An apparently general, isolated, explicit, understood action: The sole decision made at the Q2 board meeting is to put $200mm into the R&D team budget, managed by the Chief Innovation Officer.
2. A specific, enchained, explicit, understood action: The CIO decides to put the R&D team on developing a production version of a new database query tool based on a recently granted software patent for a new method of storing and searching in-table metadata.
3. A general, enchained, implicit, poorly-understood action: Because of the decision to use the new method of storing/searching in-table metadata to develop the new database query tool, investment of effort in developing existing query methods and data storage structures declines though not through explicit agreement by the R&D team.

When actions seem independent, it’s because we aren’t aware of unintended effects of an action on both outcomes and on other actions. Any broad, general action will probably produce a swathe of apparent and non-apparent results and affect other actions in apparent and non-apparent ways.

Outcomes

Like actions, outcomes can range from results which are small and trivial (the banana falls to the ground) to large and consequential (the accidental discovery of Cuba and the Americas by the Europeans, the development of a new approach to data storage and querying).

When outcomes seem isolated it is because we don’t yet understand them, and particularly how they are enchained. Take, for instance, the outcome of the broad, explicit actions to bring about the Green Revolution in the mid-20th century — globally widespread deployment of process and chemical technologies intended to increase crop yields by massively intensifying agriculture. The immediate and explicit outcome at the time was massively increased food production. (Norman Borlaug, a chemist whose work was foundational to the Green Revolution, received the Nobel peace prize for his contributions to reducing hunger.)

Over time, however further outcomes of the Green Revolution became apparent, such as declining soil fertility, changes in what people ate leading to poor health, loss of indigenous adapted crop varieties, illness in farming communities from occupational exposure to agrochemicals, and broader issues of pollution (e.g. agrochemical runoff), and environmental degradation (e.g. groundwater depletion). In a sense, these outcomes existed but remained implicit and not-yet-understood until recently. Though they were not initially overt, they nonetheless were a part of the chain of outcomes resulting from the actions associated with the Green Revolution.

Another way to say it: when we interpret a specific outcome as overtly well-understood and isolated, it is likely to instead be an outcome that is more complex. Any given interpretable outcome is connected to other outcomes in more complex ways than are immediately apparent or understandable. These outcomes and interconnections gradually become visible and understood over time.

Habitual ways of thinking about actions and outcomes

To exist in the world means taking actions and experiencing outcomes big and small, from the general to the specific, isolated or part of a sequence, apparent or not, and more or less well-understood. Everything we do and experience is unavoidably part of this soup of actions and outcomes; life both happens in and produces this environment of actions and outcomes.

Two observations:

1. We default to thinking about actions and outcomes as independent, overt, and understandable. The reality is usually that actions and outcomes are enchained, partly invisible, partly not-well-understood.
2. We prefer to take actions and obtain outcomes that are big and specific. These types of actions and outcomes are obvious and legible. But small, apparently generic actions and outcomes are more common and often (always?) interconnected with the big and specific ones.

In other words, we’re used to thinking of actions and outcomes as being simpler, less textured, and more legible than they are in reality. It’s from this gap between habit and reality that not-knowing about actions and outcomes arises.

Not-knowing about actions

Not-knowing about actions is when things you can do (i.e. actions you can take to achieve particular outcomes) are only partially known. The concepts of existing technology (things that let us do stuff, including both physical objects and manipulations of ideas and objects) and affordances (possibilities for action the agent can perceive) help in unpacking this type of not-knowing. When not-knowing about actions is resolved, the result is new action possibilities. There seem to be three sources of not-knowing about actions:

1. Not-knowing what affordances are possible from new technology: These arise when new technology — materials, processes, methods of manipulation, ways of thinking — are invented. Invention is the clearest pathway to new action possibilities because it creates new affordances. Examples include cryptographic theory for a peer-to-peer money system (2009); a method of manufacturing nonwoven fabric from ultrahigh molecular weight long-chain polymer materials like Dyneema Composite Fabric (1990s); polymerase chain reaction (1983); the Flemish technique with oil paints (15th century). Resolving this type of not-knowing about actions results in brand-new actions.
2. Not-knowing the full range of affordances from existing technology: These arise when a technology that is normally used to produce one outcome is found to produce a different outcome. Redeployment entails the discovery of new affordances for existing resources. Examples include: Using a car engine to cook food; off-label uses of approved drugs; using biological structures to design hardware or software systems like structurally coloured materials or learning classifier systems. Resolving this type of not-knowing about actions results in redeployed actions.
3. Not-knowing the contingent affordances from existing technology: These arise when a technology that produces an outcome in one situation is found to work in another situation. Recontextualisation entails the discovery of new contingent (context-specific) affordances for existing resources. Examples include: Overseas university admissions strategies applied by students from a country where all higher education is domestic; the apprenticeship mode of joining a guild applied to hiring new members in R&D teams; industrial food hydrocolloids applied by chefs in cutting-edge cuisine; longstraw heritage wheat farming techniques applied by regenerative farmers. Resolving this type of not-knowing about actions results in recontextualised actions.

Each of these forms of not-knowing about actions implies a different set of strategies and tactics for resolution.

Sources of not-knowing about actions.

Not-knowing about outcomes

Not-knowing about outcomes is when results (i.e, states of the world you might achieve through actions you take) are only partially known. This type of not-knowing seems to largely be about the intersection of imagination (can such an outcome be conceived of at all) and feasibility (can that outcome be achieved with present technology). There seem to be four sources of not-knowing about outcomes:

2. Outcomes that are already feasible and imagined somewhere else: We tend to only see and recognise these when imagination is realised and we either seek them out or are forcibly exposed to them. Examples include: national healthcare free at point of delivery; the first deployment of a weaponised nuclear reaction; the launch of a consumer interface for a large language model. However, this type of not-knowing also characterises some of the work of R&D and product teams trying to productionise and productise basic research. Examples include: developing a consumer-facing product built on top of a large language model; building the first fully curtain-walled high-rise building (900-910 Lake Shore Drive). This is probably the lowest-hanging fruit.
3. Outcomes that are not yet feasible but are imagined somewhere else: The obvious examples are from speculative fiction (e.g. warp drive from Star Trek or many of the plot elements in William Gibson’s books). However this type of not-knowing also characterises the work of some basic research oriented R&D teams and policymakers. Examples include: imagining a full multitouch phone interface and developing workable actual UX for such a device (the first iPhone); the creation of the US Constitution and its subsequent amendments.
4. Outcomes that are not yet feasible and imagined nowhere else: This is the work of speculative fiction in general, and examples aren’t possible (obviously!).
5. Outcomes that are already feasible and imagined nowhere else: This is the work of what might be termed empirical speculative fiction (such as hard science fiction).

Each of these forms of not-knowing about outcomes implies a different set of strategies and tactics for investigation and resolution. More to come on this, but to give an example: the first type might be investigated/resolved with surveys of existing outcomes (e.g., a “fact-finding trip” to a country which has figured out some desirable outcome, like a highly effective math education system) while the second type might be investigated with empirically grounded speculation involving researchers and engineers (e.g. of the sort design research companies ostensibly do a lot of).

Sources of not-knowing about outcomes.

Possibilities from not-knowing

Not-knowing about actions creates possibilities for new actions that achieve existing or new outcomes, while not-knowing about outcomes creates possibilities for new outcomes that existing or new actions can be aimed at achieving. New actions and outcomes emerge from resolving not-knowing — and the different sources of not-knowing about actions and outcomes suggest different strategies and tactics for their resolution.