Algorithmic Implications and Human Impact

Timothy Jackson and I discussed Dan McQuillan’s article “Data Science as Machinic Neoplatonism” (2018). McQuillan should be praised for his prescience in acknowledging the dangerous potentials of algorithmic discrimination and the evasion of due process. He is exactly right when he says that data science is not just a neutral new scientific methodology: it is an organizing idea—a paradigm—with cultural implications, because science itself is a cultural activity. It is historically situated, and there is no such thing as a view from nowhere. I applaud his call for a counterculture of data science that is material as well as discursive. We need more than rhetorical critiques; we as citizens need to develop our own practice of data science to counter state and corporate deployments.

Matt: "So just—I mean just to tease where I want to go—I think what he's criticizing and referring to as Neoplatonism is actually more like neo-gnosticism... because then we can see how like Plotinus has very significant criticism of the Gnostics for making matter just evil and the body just evil. And you know, I'm going to talk about Iamblichus and the importance of materiality as a sort of creating a home for the gods, that the material world is not fallen and irredeemable. Far from it, we need it to actually access the divine through ritual and through, you know, material, semiotic, material discursive practices. And so there's so much in the Neoplatonist—if you actually do a closer reading of the texts—that is saying exactly what Mcquillan is arguing for in terms of a countercultural data science."

But rather than blaming Neoplatonism for the dualistic metaphysics that informs onlooker data science, we could instead retrieve a countercultural Neoplatonism. It is true that the popular vulgar understanding of the Platonic tradition is of a two-world metaphysics, but this reflects a very poor reading of the Platonic dialogues and of the corpus of commentaries and constructive developments by Plotinus, Iamblichus, and Proclus—not to mention the alchemical tradition, which relates to matter in a very different way.

McQuillan does admit that the philosophical typology known as Platonism is distinct from arguments about what Plato himself may have believed. My wider point is not so much that he unfairly attributes views to Plato, but rather that he describes Neoplatonism in a way that resembles neither Plato’s teachings nor those of the ancient Neoplatonists. Or at least, he flattens what is actually a far more nuanced and interesting vision—not a dualism, but a continuum between intelligible and sensible, spiritual and material. The cave is a ladder, not two sealed realms.

Matt: “…the idea of historicizing science, if you recall that Timaeus begins with an account of ancient Atlantis and influence on the Egyptians, and how their understanding of number and geometry is what influenced the Greeks. And so you know, before Timaeus can get going, offering a cosmology, they have to situate their capacity to even know what number and math are about in this historical process, you know, of inheriting an understanding from prior civilizations. It's also a place-based understanding of knowledge, even, you know. And so there's a demonstration, I think, in the very structure of the narrative, in the dialogues, that scientific knowledge is culturally mediated, historically situated and place-based. You know, it's all there."

We must critique and avoid the myth of neutral facts. Meaning is not an object of uninterpreted sensation. Here again, when McQuillan says this, he is cashing a Platonic check without realizing it. Plato says the sensible realm is an image of the intelligible realm and must be understood through ideas; meaning is not simply given in the way naïve empiricists imagine. In this sense, McQuillan himself is the true Neoplatonist.

There is, in Plato—or in Iamblichus, Proclus, or Plotinus—no emphasis on the idea that the mathematical order is hidden. The mathematical order is the harmony and proportion of the visible world itself. When data science points to obscure but nonetheless predictively potent nodes hidden in the middle layer of neural networks, it appeals to a hidden mathematical order, but that order is not at all Platonic in conception; it is more Gnostic in orientation.

We must be very careful about forcing mathematical fit, because statistical correlation is not the same as causal or physical explanation. We need to maintain close contact with human forms of reasoning, which are always rooted in the sensible, embodied, culturally mediated encounter with the living world that we participate in. We are not merely onlookers; we are in and of the world. If we allow machine-learning systems—these artificially intelligent forms of computation and algorithmic violence—to make decisions for us, we give up our freedom, our due process, and our right to understand the reasons by which and for which we live.

There is a difference between logos and logistic regression; human logos is not machine logistics. We cannot relinquish reasoning, decision, and judgment to a hidden layer of neural-network nodes—only human beings can make judgments. Data science may imitate the scientific method by transforming empirical data into patterns of regularity that offer predictive power, but because we cannot understand how the nodes arrive at those patterns, we explain nothing; we merely gain a degree of instrumental power over what we do not yet understand.

The values informing our instrumentalization—informing our actions—pre-exist the algorithms and shape them. These technologies are not neutral. If the data itself embeds social bias or prejudice, then the nominally neutral output of the algorithms will embed those biases too. Underlying data reflect deeply rooted racial, gender, and cultural biases. The predictive power of these algorithms leads to pre-emptive decision-making—pre-emptive penalties and punishments framed as preventive measures—and this undermines due process, making a mockery of the idea of justice.

We cannot reduce politics to a technical operation that machine-learning algorithms could make more efficient. Politics requires human engagement; if we do not understand why algorithmic decisions are made, we will cede the practice politics and find ourselves instead living in a police state.

Plato’s "Divided Line” analogy in the Republic has a pedagogical purpose. When we first look at the sensible world, we think it contains its own meanings, but really we are just parroting common opinions; gradually we realize that our conceptual organization sorts the welter of data and extracts what is relevant to our purposes. Concept and percept are intimately entwined. Plato’s Divided Line analogy is an invitation to initiation, to a spiritual metanoia that helps us recognize the role of conceptual activity, not to escape from the sensory world, but to reveal its meaning. We need not only individual epistemic techniques but also social apparatuses—what Foucault calls dispositifs—which Plato implicitly acknowledges in the Republic.

Thus the very idea that facts are constituted by an overall pattern or gestalt—that “facts are made by frames”—is itself a Platonic insight. Likewise, tying knowledge to justice (for instance, through the concept of “epistemic injustice” highlighted by McQuillan) is straight out of the Republic, where Plato offers both the Divided Line for epistemology and a picture of the just city and the just soul. There is no veil of neutrality in Platonic thought: knowledge must be oriented toward the Good, and truth cannot be divorced from value. Plato’s epistemology is aesthetic and axiological from the start. The point is not to pretend neutrality but to own our situated, value-laden stance, just as feminist standpoint theory (eg, Sandra Harding, Donna Haraway) urges.

Matt: "Because I think that the problem is, we tend to understand the eternal. We reify the eternal as an origin in time. When really it's the Platonic conception of time as a moving image of eternity and the Whiteheadian conception of concrescence. Where and you know the whole mystical tradition. Going back, you know, to Eckhart and others, is that, like the Eternal is fully present now, you know it's not the opposite of time. It's the—it's the plenitude of time. It's the superlative form of time. It's the wholeness of time that's present in each of the instances of time."

McQuillan notes that Copernicus turned to the Pythagorean astronomers to help stimulate his articulation of the heliocentric model of the solar system. Crucially, however, Copernicus’s heliocentric model was actually less predictively accurate than Ptolemy’s geocentric system. The historical commitment to the Copernican view therefore has nothing to do with the triumph of empirical observation over inferior superstition. Instead, Copernicus inaugurates a whole new organizing idea: although inferior in predictive accuracy, his model was perceived as superior because it built on geometric symmetries. It was simpler than Ptolemy’s epicycles, even though it did not align as well with observations. Thus, symmetry—not data—guided Copernicus’s adoption of the model.

Similarly, Galileo re-defines motion in a way foreign to Aristotle’s physics. For Aristotle, motion concerned something essential to the body itself—a purposeful unfolding, a developmental process immanent to the object. Galileo realizes that motion is indifferent to any absolute state: even though the whole Earth is moving, bodies on Earth share its inertial frame, and thus they seem at rest. This abstraction leads directly to Newton’s concept of inertial motion and marks a decisive break from the older, phenomenological understanding of change.

Plato’s Timaeus provides an early model for historicizing science. The dialogue begins with an account of the ancient Atlantean civilization, from which the Egyptians, and then the Greeks, inherited their understanding of number and measure. In situating their inquiry historically and geographically, the interlocutors demonstrate that scientific knowledge is always mediated culturally and place-based.

McQuillan rightly calls data science a form of rhetoric rather than an X-ray of reality. Karen Barad’s agential realism underscores this: science breaks with classical representationalism, revealing itself as a human making—an art as much as a discovery. Thus, science is creation as well as revelation, and, as McQuillan argues, data science must be recognized as such.

Tim also mentions this paper by Huh et al.: “The Platonic Representation Hypothesis”