Lecture 6-2: The Lens of Biblical Justice — Mishpat and Tsedaqah

Micah 6:8 — "He has shown you, O man, what is good. What does Yahweh require of you, but to act justly, to love mercy, and to walk humbly with your God?" (World English Bible)

Amos 5:24 — "But let justice roll on like rivers, and righteousness like a mighty stream." (WEB)

Deuteronomy 16:20 — "You shall follow that which is altogether just, that you may live, and inherit the land which Yahweh your God gives you." (WEB)

3.1 Two words, not one, and why they travel together

Mishpat (מִשְׁפָּט) is most often translated "justice" or "judgment." Its root sense is forensic and administrative: the act of a judge rendering a verdict, or more broadly the ordering of a society's institutions so that verdicts are rightly rendered. When Amos or Isaiah accuses Israel's leaders of failing at mishpat, the charge is specifically institutional — courts that take bribes, judges who favor the powerful, a legal order that no longer does the one thing a legal order exists to do, which is protect the vulnerable from the strong by applying the same standard to both.

Tsedaqah (צְדָקָה) is usually translated "righteousness," but this English word can mislead a modern reader into hearing something narrowly individual and moral — personal piety, private uprightness. In its Old Testament social usage, tsedaqah is closer to "right relationship" carried out concretely: a righteous person or community is one whose relationships — to God, to neighbor, to the poor at the gate — are conducted rightly, which in practice means with active care for those the relationship could otherwise be used to exploit.

The two terms are paired often enough — Amos 5:24's parallelism ("justice... and righteousness") is a clear instance, and Deuteronomy 16:20's near-redundant "that which is altogether just" (literally closer to "justice, justice you shall pursue," the doubled word itself an emphatic construction) is another — that Old Testament scholars generally read the pairing as a hendiadys: two words functioning as one compound idea, comprehensive social justice, rather than two separable virtues that could in principle be pursued apart from each other. This matters directly against a possible modern misreading that would split them into "structural justice" (mishpat, an institutional matter) and "personal righteousness" (tsedaqah, an individual matter, safely private) — the text does not license that split. A community can no more have tsedaqah without mishpat than a court system with fair procedures but unjust outcomes could be called just.

3.2 The recurring beneficiaries

The prophetic justice texts do not leave "the vulnerable" as an abstraction. A short list of named categories recurs with striking consistency across the legal and prophetic material: the poor, the widow, the fatherless (orphan), and the sojourner (the resident foreigner without the native-born person's standing or protection). These four groups share a structural feature rather than only an economic one: each lacked the ordinary social leverage — a functioning kinship network, land, citizenship status — that protected an Israelite from exploitation by default. The law's justice provisions are aimed disproportionately at exactly these groups not because the text plays favorites among sinners, echoing Leviticus 19:15's warning against partiality in either direction, but because a society's justice can be read most reliably at its margins. A court that treats the powerful fairly and the powerless carelessly has not achieved partial justice; on the biblical account, it has failed at justice, because justice was never primarily being tested on the cases that were already going to come out fine.

This has a direct and non-obvious application to the algorithmic case. A system's aggregate accuracy statistic — "ninety-seven percent overall accuracy" — is precisely the kind of summary that can look excellent while concealing catastrophic failure at the margins, exactly where Gender Shades found it (lecture 6-1): overall accuracy was respectable; accuracy for the least powerful, least represented group in the training data was not. The prophetic instinct to ask "how does this treat the widow, the sojourner, the one without standing?" rather than "what is the overall rate?" turns out to be, among other things, good statistical practice — it is the instinct behind testing a model's error rate by subgroup rather than only in aggregate, arrived at from an entirely different direction than a fairness engineer's dashboard, but landing on a strikingly similar diagnostic question.

3.3 Contemporary theological voices

The argument that biblical justice is a whole-society category, not an individual virtue alone, has serious contemporary defenders worth engaging directly rather than assumed as settled. Nicholas Wolterstorff's Justice: Rights and Wrongs (2008) develops a philosophically rigorous account of justice as grounded in inherent human worth — a worth that generates real claims (rights, in his vocabulary) that others are obligated to honor, rather than justice being merely a virtue a generous person happens to practice. Wolterstorff explicitly engages the Old Testament justice texts as a central resource for this account, not an illustration added after the philosophical work is done.

Tim Keller's Generous Justice (2010) makes a more pastorally oriented version of a similar case, arguing from the same mishpat/tsedaqah pairing that biblical justice is inseparable from active care for the economically vulnerable, and that this is not a peripheral social-justice add-on to the gospel but flows directly from it. Keller is worth naming here specifically because his audience and rhetorical register (evangelical, pastoral, not academic-technical) demonstrates that this reading of mishpat and tsedaqah is not confined to a particular theological wing; versions of it are held across a wide range of traditions, even where they disagree sharply on other questions, including — worth flagging honestly — real disagreement within that same broad evangelical world about how directly biblical justice language should be applied to contemporary structural and political questions. This course takes the position, following Wolterstorff and Keller, that the application is warranted; a careful student should know the position is not universally shared even among readers who accept the same texts as authoritative.

3.4 Applying the lens directly to an algorithm

The move this lecture wants a student to be able to make is not "algorithms should be fair" — a claim thin enough that almost no one would state the opposite — but the more specific move of applying mishpat and tsedaqah's particular content to a system's output.

Take a hypothetical: a hospital deploys an algorithm to identify patients who would benefit most from a limited enhanced-care program, using healthcare spending as a proxy for how sick a patient is (a real design pattern documented in the fairness literature, notably in Obermeyer et al.'s 2019 study of a widely used commercial risk-prediction algorithm). Because Black patients have historically had less access to and lower average spending on healthcare even at the same level of underlying illness — itself a structural inequity — the algorithm systematically underestimated how sick Black patients were relative to white patients with the same actual health needs, and referred them to the enhanced-care program at a lower rate. Read through a bare "fairness" lens, one could argue the algorithm treated the input (spending) identically for everyone. Read through mishpat and tsedaqah, the question is not whether the rule was applied consistently; it is whether the outcome served or failed exactly the population — the person with less access, less standing, less historical claim on the system's attention — that the prophetic texts insist justice must be tested against. By that standard, an algorithm that reproduces the disadvantage of the already-disadvantaged has failed mishpat, regardless of how consistently its internal rule was applied, and a church or hospital ethics board with this vocabulary available has a sharper diagnostic question to bring to a vendor than "is it fair?" — namely, "who is this decision-rule serving worst, and does that match the group Scripture insists we test justice against?"