Against Occam's Razor

On Pitfalls of 'Rationalist' Thought, Positive and Negative Definition, and Models of Reality

Occam's razor (aka the 'law of parsimony') is not, as is commonly thought, a principle of rationality (much less a "law"). The principle or law of non-contradiction, advanced by Aristotle and others, is an example of just such a fundamental principle of rationality. Occam's razor, however, is more or less just a heuristic, a tool, and a useful one but nowhere near as universally applicable or valid as is commonly assumed. A thought experiment to illustrate the point: if you were being investigated for a crime that you had not committed, but through unlikely circumstances, you happened to appear to be the most obvious suspect, certainly you wouldn't want the detective or judge to convict you simply because the true perpetrator committed a crime in an unlikely, non-obvious, or convoluted manner.¹

Occam’s razor itself is not necessarily the issue. Instead, the flaw here is a type of thinking that is safe, and calculating, but ultimately lacking in creativity. Occam’s razor no doubt has a range of usefulness but can become quickly overbearing and stunt investigation when it becomes facilitated inappropriately by a domineering mode of thinking that reduces rational thought to a grab-bag of heuristics and mental tools. This thinking proliferates more as the aesthetic of careful thought, an image of what it means to be a 'rational' thinker, while the real substance of critical thought lags behind. It is common among internet ‘rationalists’ as one example,² but the effects are wide-reaching; arguably it grips Western post-Enlightenment thought as a whole. Facilitated by this type of thinking, such heuristics are abused, overused, and overextended, as will be explored below.

More notably, the fundamentally reductive character of this thought stunts and warps our worldviews; flexible, nonlinear thinking is devalued, and analytic, linear thinking is granted supremacy. The pitfalls of this dynamic are the root of some glaring modern issues that will be explored here and in further articles.

What I aim to illustrate here is not that parsimony isn't useful, but rather to explore the implications of this dominant, shoddy thinking that over-relies on parsimony at the expense of addressing the complex nuances and subtleties essential for an accurate and comprehensive understanding of reality.

Some Issues in Parsimonious Thinking

Unlikely things do indeed happen: the value of statistical and causal thinking

As noted, the overuse of this heuristic leads to rather circumscribed and inflexible approaches to thought. The thought example provided before illustrates that parsimony alone cannot resolve whether or not something unlikely or counter-intuitive is the real reason behind a certain event.

When weighing the likelihood of whether some rare event may have occurred, it’s worth remembering that just because something is rare does not mean it may not have happened (cf. ‘What Makes Statistical Thinking Useful?’), even in ways that appear 'absurd', and especially those that occur in non-obvious ways that defy conventional logic.

Indeed, when one dives into the study of causality and causal inference, one thing becomes abundantly clear: basic human causal reasoning, as surprisingly powerful as it may sometimes appear, is all too often devastatingly wrong. The research sciences are replete with examples of highly counter-intuitive causal pathways that conventional reasoning failed to discern, and were ultimately only realized via entire programs of research and the consolidation of bodies evidence across multitudes of investigations. Overconfidence in simple heuristics should instead be replaced by a flexible, epistemic humility, one that motivates us to avoid slipping into tempting, but misleadingly simplistic assumptions. This epistemic humility also leads us to recognize the need to draw from a range of complementary bodies of knowledge, rather than become locked into the parochial confines of one or another specific discipline or way of thinking.

a range of usefulness

Another issue concerns the overextension of parsimonious thinking outside its range of usefulness. Simply put, as a carefully designed heuristic, it does not claim nor merit application outside of very particular domains of decision-making. When might it be useful then? When comparing two possible, highly abstracted conceptual models, one with many premises and another with fewer, will parsimony be relevant? Probably.

But consider complex sociohistorical realities. Will you be able to Occam’s razor your way into an accurate historical understanding of Apartheid South Africa?

What about complex moral quandaries, are they served by such formulaic thinking? In the face of grave societal injustices, is amnesty the definitive approach, or retributive justice? There’s no Occam’s razor for whether one should aim to preserve societal stability or to preserve the rights of victims in the aftermath of injustice (whether of apartheid, legacies of colonialism, the aftermath of genocide, and so on). And yet we find ourselves at a moment of history where answers to just such questions are desperately needed, and careless thinkers apply rabidly simplistic solutions to these social and moral challenges, a symptom of overcommitment to this heuristic mode of thinking.

These are not questions we can answer simply (or on our own), and certainly not with a heuristic. In other words, our basic mental tools and approaches to understanding some phenomena need to fundamentally match the nature of what we wish to understand (more on this later). The tools required to adjudicate between two scientific models, and two paths of reaching social harmony are categorically different, as these are categorically different realities; yet this dominant reductive form of thinking I've referred to above tempts even otherwise bright and capable minds to collapse these complexities into questionably simplistic formulas.

easy thinking, negative and positive definition

As the ‘razor’ in the name suggests, Occam’s razor is an example of a type of thinking that gains legitimacy from the act of tracing out the ‘negative' or ‘unknown’ and cutting away context (e.g. abductive reasoning). Occam's razor thrives when superfluous information clouds judgment about core principles, or when an abundance of constructs overburdens our conceptual understanding with too many features to grapple with effectively.

The useful solution is to cut those away.

The remainder has been defined by exclusion, or negatively defined. Such negative definition is applicable in many circumstances. For instance, instead of advancing a claim, this approach may delimit another claim by pointing out a lack in it, a flaw or otherwise attempting to refute or falsify it.³ It is not uncommon that certain claims that are themselves unfeasible or impossible to prove, may be supported if you can at least disprove (mutually exclusive) alternatives. Refuting alternatives lends evidence towards such claims.⁴

Negative definition stands in contrast to the much more difficult (and ambitious) task of positively articulating and defining the reality or presence of some thing or advancing an explicit proposition or claim related to it. For instance, when you advance a certain cause as the origin of an effect, this is an act of positively definition. When you propose a theory that provides explanatory power, or create a construct to describe something (say, agreeableness in psychology), you are positively defining something. For more on the nature of this dialectic of ‘positive’ and ‘negative’, see the following post:

The asymmetry of negative and positive definition

Both negative and positive means of definition are extremely useful and potent tools. However, there is a dynamic asymmetry between them. All too often defining something in the negative, solely by cutting away context and higher dimension complexity, ends up being a type of ‘easy’ thinking that never challenges you to positively articulate the broader complexity of the matter at hand. This dynamic is one thing that makes critical thought so definitively useful and yet seductive. In the act of critique, you emphasize pitfalls and failures of other propositions, you strip away the faulty or false and reveal (hopefully) the truth that remains. At its extreme, one may never even tie themselves strictly to any particular premise or way of thinking that requires adherence to a positively advanced claim, worldview, or coherent system of thought. One needs only to retreat to the safety of deconstructing or stripping away apparent flaws of another proposition, theory, or worldview.⁵

The effects of this powerful asymmetry in thought have a significant negative impact on the quality of intellectual life in our society. Serious examples of this abound.

For instance, consider the phenomenon of 'NIMBY'. Why has such an obstinate behavior, observed in so many cities, proliferated almost automatically, without any central effort or mass organization? This is partly due to the asymmetry between negative and positive definition. It is far easier to coalesce around the rejection of some change (in favor of the status quo) than it is to coalesce around a novel positively defined change. This leads to a kind of ‘black hole’ or ‘gravitational’ effect; we have too-few thinkers in the realm of boldly, positively staking claims and too many in the miserly camp of stripping away things or seeking any means to tear something down or reject something without ever advancing something of themselves. It is simply too easy.

Far too many gravitate towards the ultimately easier task of 'minimalist' thought or the similarly seductive realm of deconstructivist critique. This is no light matter to overcome, and requires a quality of mind, a fluency in a particular mode of thinking—a synthesizing, integral approach—that is tragically underdeveloped in our society or, perhaps, in human cognition in general.

health, illness and the moral domain

To get a sense of how omnipresent this dynamic is, consider the same asymmetry between defining disease and defining health. In one sense or another, the last four centuries of medical advancements can be attributed to evermore specific and complex taxonomies of disease. But what about what it looks like and means to be healthy? Is there any such similarly comprehensive 'taxonomy' of what it means to be healthy? Is health merely the absence of disease, as this state of affairs may lead us to believe? What is lost in only ever attending to one side of this dynamic?

If you want to see this phenomenon in action, go review some scientific publications in the health sciences. Observe the extraordinary variety of operationalized metrics to define disease states and statuses. You'll notice that the dominant framing is one of lack, or rather of emphasis on the diseased state and its manifestations. But what about what constitutes health? Such metrics and research operationalizations are present, sure. But in far fewer numbers.

Another arena where this dynamic may be seen is our conceptions of equality and inequality, or justice and injustice. In other words, the moral and intellectual domain of humanity. No matter the variance in opinions, one will likely find that reaching a consensus on what constitutes an inequality or injustice is actually rather forthcoming. Counter to what you might think at first, it is the 'negative' that is more apparent because the brute reality of oppression or unequal treatment is visceral and evident.

But so soon as you transition into the realm of defining what equality or justice are or what they look like, the matter becomes far more difficult and complex. Indeed, depending on context and culture we may find that something that is unequal or unjust in one context, is just or equal in another, because of differences in worldview, models of reality, morality, and the like.⁶ Recognizing this demands a fundamental shift in our capacities. To engage in the act of positively investigating equality and justice effectively—let alone collaboratively and socially, in a way that avoids parochialism, ethnocentrism, reductionism, or relativism—is incredibly demanding.

To illustrate the asymmetry, we can conceive of someone who spends their entire life finding examples of injustice and gets quite good at it, and yet never takes the step of practically thinking through what 'Justice' actually is, or exploring what it means to experience it, or to contribute to a society-building movement towards bringing about justice.

Returning to the prior questions about amnesty and retributive justice, can we ever hope to make substantive progress if we systematically deprive ourselves of opportunities to positively define the qualities and characteristics of equality and justice? qualities that are undoubtedly necessary if we are to make progress in such questions?

An Aside on Models of Reality

Part of the confusion with applying principles of minimal assumptions relates to explicitly thinking about what models one adopts and how one conceives of models more generally. Without delving more comprehensively into the nature of 'models of reality', we can say definitively that parsimony is a very useful principle when working with discrete (and similarly likely) models. As alluded to before, in practice, adhering to the more simple model among a group of models is a productive heuristic because a model's usefulness derives from its broad applicability and simple flexibility for explaining observations and deriving sensible conclusions, rather than how exhaustively and precisely it fits reality.

In fact, it is useful to think about models generally as always already being wrong (as in there is no absolute correspondence with reality - because this cannot be so). Ideally, they are usefully wrong. Thus, the value of simplicity is in refining the formulation of the model by redefining unsuitable premises into more broadly applicable and encompassing ones, and retaining its core function as a useful, circumscribed depiction of reality. This can be seen in examples of models that describe evolutionary phenomena, or conceptual models of other complex phenomena, from models of the social determinants of health, to demography and flows of population movement and migration.

However, reading the above carefully, the idea that simplicity is a value is critical because it reveals that as much as parsimonious thought often leads to credible and useful conclusions, it is also as much an aesthetic or style of thought. Indeed, in our post-Enlightenment culture that vaunts rationality, certain types of thinkers gravitate towards this way of thinking; our culture lauds it, reinforces and proliferates it, and overlooks its hazards. To expand on this further is a subject for another day.

Conclusion

In sum, the challenge here isn't whether or not parsimonious thought and Occam's razor are useful mental tools, they certainly are. Instead, the challenge comes from overusing and overextending a certain tool beyond its sphere of usefulness or crafting an entire intellectual identity from what is ultimately a style of thought among many valid styles.

The danger comes from univocally applying the principle to all domains, including domains that don't merit it. This is especially apparent when reaching into areas of reality, e.g. social phenomenon, that are irreducibly more complex than natural systems—which, of course, are already absurdly complex in their own right.

When it comes to complex social realities, there’s no suggestion that the principle of minimalism will enable us to finally come up with a definitive scientific model for the sociology of religion, for instance, or for solving global malnutrition (or reconciliation after apartheid, as referred to before). Indeed if it were applicable to such arenas, perhaps we might have expected better progress on those fronts given how many great scientific minds are using Occam’s razor in practical, circumscribed contexts. And yet neither the rationalists nor the functionalists — not to mention other interpretive models — seem closer to the definitive definition of religion, or of restitution and reconciliation for that matter, let alone the many other complex domains of human life that are no less resistant to circumscribed parsimonious thinking.

1. If this seems fanciful, one need only consider the case of Sally Clark and the errors in reasoning that led to her conviction to find ample reason to consider this seriously.

2. ‘rationalism’ and ‘rationalists’ of the internet era - what I mean by 'rationalism' here is a loose collective of pop intellectuals that have emerged in public discourse throughout the 2000s and 2010s. Bringing together New Atheism (Sam Harris, Richard Dawkins), techno-libertarians, alt-right, with a heavy dose of the Bay Area trans-humanism & tech-utopianism of Silicon Valley... Not necessarily just the “thought-leaders” in these spaces, but more so the group of individuals and common culture they’ve created, the style of thought, and the ways the ideas they are attracted to and promote spread and affect others in the landscape of contemporary thinking, at least on the internet.

3. Indeed, such ways of thinking became the raison d'être of mid-century epistemology of science, particularly within the framework of ‘falsificationism’.

4. For more on this, cf. ‘proofs by contradiction’ and ‘disjunctive syllogisms’.

5. For a superb exploration of this phenomenon in relation to ‘critique’, see Bruno Latour’s essay Why Has Critique Run out of Steam? From Matters of Fact to Matters of Concern.

6. Note, that this is not an endorsement of moral or prescriptive relativism. Although the room is not available to explore further, this observation is intended to fit alongside a careful and considerate cultural and descriptive relativism without granting that these realities compromise any hope of consensus and clarity about reaching truth and reality.