This article presents a particular method of analysis and family of social theories I find important. It also builds on the concept of expected value.
I. Jane and The Canyon
Jane is an explorer working her way through a forest. She has gear, food, and a general direction to travel, but the obstacles she will face and the paths she could take are shrouded in mystery.
This morning Jane found a stream flowing in the direction she is headed. Since water is neither ubiquitous nor optional, and the rocky “banks” of the stream are easy to walk along, she decides to follow it. As more creeks flow into it, the stream becomes a river that begins to carve the landscape. With every step the now-river is bordered by taller banks, but there is still plenty of space for Jane to walk. Given the choice between a steep climb up the dirt that defines the river or an easy hike along the river, she chooses the latter. Slowly the banks of the river shrink until Jane looks ahead and sees the cliff faces close in and for a few feet there is nothing but river, after which the dry banks of the river return in full. The cliffs will take a while to climb up (and longer to safely descend) so Jane chooses to wade forward before returning to dryland.
Jane continues until the canyon narrows and she has to wade again. By the fourth time this happens she decides to walk with the current—it is faster than taking off her shoes every five minutes. In midafternoon Jane experimentally tries wading against the current and is surprised by how difficult every step is. It will take a while, and a lot of energy, to backtrack to the earliest camping spot she passed, but there is a possibility that forward progress will provide more land. Jane presses on until she reaches a series of small waterfalls.
None of them are that large and Jane figures she can climb down safely, but climbing back up with her gear will be significantly more dangerous and difficult (and there is no space to store her gear above the waterfalls). At the bottom of the waterfalls there is some dryland, but after that the cliffs narrow in on the river with no end in sight—and at no point do they seem to shrink. As difficult as climbing up the cliffs to the forest above the waterfalls would be, at the bottom it will be impossible. Some clouds have moved in and while the day is not yet over, the canyon is deep enough to promise a fast sunset. Jane is in a dangerous situation. As she weighs her options, she asks herself. “Where did I go wrong?”
II. Path Dependence
In the social sciences there is a perennial debate around agency and structure. Loosely, agency is an actor’s ability to make her own consequential decisions and create her own future while structure is the degree to which outside forces (such as laws or cultural norms) determine actors’ actions and drive history. Extreme forms of structuralism, such as pure Marxism, leave little to no room for individual agency. Meanwhile some theories, like libertarianism, see human agency as the determining factor in everything, with what others call structures being either non-existent or simply an aggregation of the agency of a group of other individual actors in this moment.
Most theorists (including some Marxists and libertarians) find some middle ground between these extreme. Personally I lean towards path dependency. Originally an economics term, in short path dependence in political science believes that actors agency is minimal except when they have a chance to change structures. This theory revolves around four key assumptions:
- When structures are weak agency is strong.
- When agency is strong it can, and will, generate structures that constrain agency in the future.
- There is room for agency except when structures are at their absolute strongest, it is just heavily constrained and directed toward a particular path.
- Movement down a particular path reinforces that path and its accompanying structures.
Jane’s story is one of path dependency: because she chose to follow the stream she faced a series of decisions, the logically correct choices of which ended with her trapped in a canyon. Had she stuck to higher ground the landscape (structure) would have determined her choices but both the options and outcome would have been different.
III. Thinking with path dependency
Early on Jane chose to follow the stream. From then on she calculated the expected value of following the stream or backtracking and followed the logical solution which pushed her forward, further locking her into the path of the canyon, leading her to decisions in the afternoon she would have balked at in the morning.
Every calculation, as structured and given the then available information, may have been accurate. The mistake was in the initial assumptions that accompanied the calculations. Any radical shift in strategy, such as turning back or climbing out of the canyon, was rarely considered as a viable alternative. To the degree turning back was considered as an alternative to the river, the short term difficulty of the choice was fused with its inherent medium and long term uncertainty, and compared to only the short term ease of following the river and medium term advantages of the river. The fact that following the river constrained future decisions did not factor into the calculations until too late.
Theory is only as useful as its application, and path dependency is no exception. If we live in a world defined by path dependency (which I believe we do), then it must be a part of every expected value calculation. This is not the “outcome X value of outcome X probability of outcome” calculation that most of expected value questions are. Rather, it has ~six components (ouch, I know):
- What structures (if any) could be strengthened or weakened?
- And what is the value of the structure in question?
- To what degree are they being strengthened or weakened?
- What is the probability this outcome does in fact strengthen or weaken the given structure to this degree?
- What is the probability that this strengthening/weakening has a notable effect on the structure?
- What is the probability that this strengthening/weakening will come in combination with other actions that collectively will have a notable effect on the structure in question?
The question of what structures are being influenced, the value of those structures, the degree to which they are being strengthened or weakened, and the probability they are strengthened/weakened collectively make up the normal expected value calculation. The issue with structures, however, is that they are not really continuous. Rather they don’t exist, they are deteriorating, they are weak, they are gaining strength, or they are strong. Changes between these categories are important, but changes within these categories have minimal importance. So an action may have a significant effect on a given structure but if it isn’t significant enough to cause a categorical change in the structure it matters only in so far as it allows future actions to cause a categorical change—the probability of which should be factored into the calculation too.
Obviously, most of the time a lot of data are missing and this calculation isn’t always possible. Likewise, it is complicated and fraught with uncertainty so while it should be a component of expected value calculations for practical reasons it is often left out. The takeaway I want to provide here is twofold: First, in moments of flux when structures are their most malleable this needs to be part of the conversation. Second, I want there to be an astrix by every calculation that does not include it, just as a reminder of what is being ignored. Otherwise I fear we as a society won’t be able to create adequate systemic change when the opportunity presents itself.
This post was getting longer than I wanted so I have moved all the examples of path dependency here as something of an appendix.
The Treaty of Versailles is a good example of actors with significant agency making decisions that locked them into a path. The Treaty could have been written in any of a number of ways and all actors involved worked to shape it to their liking. It was not guaranteed that the Treaty would impose harsh penalties upon Germany, but the signed document did impose those penalties, nearly guaranteeing that Germany would come into conflict with France and other Allied powers in the future. The continued punishment of Germany as its economy weakened further pushed the world towards conflict. Had Hitler’s Nazi Party not started a war, another group would have.
This topic well understood in the international security realm where questions like “if Turkey invokes Article IV of the NATO charter against the Kurds should the US respond?” often contain an argument akin to “doing anything other than supporting Turkey will destroy NATO” along with the pros and cons of the conflict itself.
Abu Ghraib provides a case of actions significantly influencing a structure (in this case, torture conducted by the US government) but not weakening it enough to cause a categorical shift. The guards were punished but because nothing further came of the crimes at that time torture programs continued and now Gina Haspel (who has been implicated in torture scandals) is now the director of the CIA. Had Guantanamo, and every other blacksite, been shut down publically or people higher in the administration (such as Donald Rumsfeld) suffered consequences torture would likely not be a component of the US intelligence community in any fashion.
Precedent in law is defined by path dependency. For example, in the 1803 Supreme Court case Marbury v. Madison the Court ruled in favor of James Madison (and by proxy Thomas Jefferson’s anti-federalists) based on the principle that the law in question was unconstitutional. In doing so they established SCOTUS as an organization capable of striking down any law based on constitutionality, changing the power of the court in a now irrevocable fashion. Then President Jefferson had to choose between accepting his victory (and allowing the precedent of judicial review to be established at the federal level) or something radical like siding with Marbury and the Federalists to undermine SCOTUS (and likely causing a constitutional crisis). Had Jefferson realized the far reaching consequences of the decision, I wonder if he would have chosen differently.
My final example is another case where actors did not consider path dependency. As the #MeToo movement picked up steam in 2017 and 2018 numerous leaders in the Democratic Party made a calculation that the movement would help them electorally (in no small part because of the Access Hollywood tape) and thus gave it greater voice. It is quite possible that had they considered how the movement would prevent them from defending individuals in their party—as happened to former Senator Al Franken—they would not have been so vocal.
 Here, and in the rest of the post, I am using “structure” in a horribly broad way to mean anything that constrains agency. I apologize to every cringing theorist.
 Since structure is a necessary condition for many forms of progress this is not necessarily a bad thing. For example, after World War II the Marshall Plan and the increased economic union between Germany and France created a structure that has led to the longest peace between those two countries in history.
 I have no doubt that had Franken been a senator in the 1990s and faced a similar scandal his political career would have continued largely unhampered. From a purely electoral calculus in favor of the Democratic Party over the next four years I don’t know where the party’s leaders pushing #MeToo falls. From a moral calculus, I am entirely in favor of #MeToo—and I think Franken’s resignation was worth it to build the social norms espoused by the #MeToo movement. Even though he was an otherwise powerful player on The Left of US politics.