Multiple sources of legal authority
Different communities accumulate their own separate bodies of case law. A choice-of-forum rule decides which body each new case is judged under.
Scroll the diagram sideways to see all of it.
aiag/cases/legal_pluralism.pyimport aiag
# The 4 approaches of Legal pluralism, with the classes that carry them out
for approach in aiag.approaches(locus="case", conflict="partition"):
print(approach)pip install aiagaiag/cases/legal_pluralism.py Every method in aiag Library on GitHub
Limitation
Users learn to exploit the routing. A case described in the right way is routed to the body of case law that gives the answer the user wanted. The choice of forum, rather than the merits of the case, then decides outcomes.
This approach takes the decisions a group has already made (the posts it removed and the moderation decisions it let stand) and works backward to the rule it was actually applying. The rule is written out in a form a person can read, either a scored table of telltale phrases or a named list of norms. This written rule is the goal of the approach. It becomes the group's operative standard, which a human or a model is then held to when the next case arrives. Because it is explicit, it can be checked and disputed. Being explicit also allows each rule to be sorted by how far it reaches, i.e., whether it is enforced platform-wide, shared by a cluster of similar groups, or specific to one group. A group's own standard can thus be separated from everyone else's. The accumulated record ends up as a rulebook, not as a classifier.
One trained system does the judging. What varies by community is how that system is adapted to the community where it runs and who receives the ruling. The papers adapt it in two ways. The first uses a set of community-specialized expert models, with an allocator that picks which expert rules on a given post and an explainer that gives the reason in that community's own vocabulary. The second trains a single model on data pooled across a hundred communities to capture what they share. It then sets a separate decision threshold for each community and refers its decisions to that community's own human moderators. The two differ in where the community is represented, either in separate experts or in a threshold and referral process on top of a shared model. Both end the same way, with one ruling for one case. In both, the community's own accumulated history decides where its line falls and who reviews the ruling.
This approach models each individual annotator, conditioned on the groups they belong to. A practitioner then states in advance which groups sit on the jury and in what proportion. The system reports the verdict that this jury would reach and how divided it was in reaching it. Nothing is routed and nothing is trained for each community. Instead, the panel is assembled on demand from individual-level models. The same case, run again with a differently composed jury, can therefore come out differently. The jury's composition is an explicit setting rather than a learned choice. Disagreement is part of the output rather than noise to be averaged away before the answer is given.
The system keeps accumulated decisions as a collection of past cases, each with its recorded reasoning. At decision time, it retrieves the cases relevant to the current situation and reasons from them in context. Nothing is trained for each group and no rule is written. Serving a new group or a newly written policy therefore means adding cases to the collection rather than retraining, and this portability is the main purpose of the approach. The two papers divide the collection differently and retrieve differently. One indexes cases by demographic group and deliberately prefers scenarios that sharply separate groups over scenarios that merely look similar, so the retrieved examples carry the disagreement. The other indexes cases by user-written policy and retrieves earlier inputs similar to the current one, together with the critique-and-revise reasoning they produced.
Several bodies of norms operate at once within one social field. Rather than one official legal system with exceptions, there are multiple active systems of rules side by side, each generating its own decisions.
Normative orders do not just sit in parallel but interpenetrate. People live at their intersections and are subject to several at once. The boundaries themselves are porous.
When several lineages of decisions are in force, a stated rule must assign each new case to one of them. Otherwise, parties simply shop for the forum that favors them, which is the failure predicted for this regime.
What Is Legal Pluralism?
1986
Legal Pluralism
1988
Toward a New Legal Common Sense
2002