This tutorial will show how to use AuthoritySpoke to model legal rules found in legislation. This is a departure from most of the AuthoritySpoke documentation, which focuses on judicial holdings.
These examples are based on the fictional Australian Beard Tax (Promotion of Enlightenment Values) Act 1934, which was created thanks to the New Zealand Service Innovation Lab, to supply test data for experimental legal rule automation systems.
The Service Innovation Lab's version of the Beard Tax Act is a PDF. AuthoritySpoke is designed to load legislation from JSON data using a related Python library called Legislice. So I've set up a web API that can serve the provisions of the Beard Act. You can also browse the provisions of the Beard Act in your web browser.
For convenience, this tutorial will use Legislice's mock API server with fake JSON responses, instead of connecting to the real API.
import json
import os
from authorityspoke.io.fake_enactments import FakeClient
legis_client = FakeClient.from_file("beard_act.json")
Next, we'll load annotations for the legal rules found in the statute provisions. These come from a JSON file that will be loaded as a Python dictionary. AuthoritySpoke rules are procedural, so they have one or more outputs, and zero or more inputs. They can also have "despite" factors, which are factors that may not support the output, but they don't preclude the output either.
AuthoritySpoke rules are usually supported by enactments such as statute sections. These can be retrieved from the API with the URI-like identifiers like those in United States Legislative Markup (USLM). In this case, since the Beard Tax Act is Act 47 of 1934, the identifier for Section 4 of the Act is /test/acts/47/4.
In AuthoritySpoke, you have to think about two JSON schemas: there's one schema for legislative provisions fetched from the web API, and another schema for rule annotations that you (currently) have to create for yourself. Of course, you can create either type of object directly in Python instead of loading them from a JSON file. For details, see the AuthoritySpoke reference manual. Here's an example of one JSON rule annotation.
from authorityspoke.io import loaders
beard_dictionary = loaders.load_holdings("beard_rules.json")
beard_dictionary[0]
{'inputs': [{'type': 'fact', 'content': '{the suspected beard} was facial hair'}, {'type': 'fact', 'content': 'the length of the suspected beard was >= 5 millimetres'}, {'type': 'fact', 'content': 'the suspected beard occurred on or below the chin'}], 'outputs': [{'type': 'fact', 'content': 'the suspected beard was a beard', 'name': 'the fact that the facial hair was a beard'}], 'enactments': [{'node': '/test/acts/47/4', 'exact': 'In this Act, beard means any facial hair no shorter than 5 millimetres in length that: occurs on or below the chin'}], 'universal': True}
The "universal" True/False field indicates whether the Rule is one that applies in every case where all of the inputs are present, or only in some cases. The default is False, but this Rule overrides that default and says it applies in every case where all of the inputs are present.
Now we can have AuthoritySpoke read this JSON and convert it to a list of Rule objects. In particular, we'll look at the first two Rules, which describe two ways that an object can be defined to be a "beard".
from authorityspoke.io import readers
beard_rules = readers.read_rules(beard_dictionary, client=legis_client)
print(beard_rules[0])
the Rule that the court MAY ALWAYS impose the RESULT: the fact that <the suspected beard> was a beard GIVEN: the fact that <the suspected beard> was facial hair the fact that the length of <the suspected beard> was at least 5 millimeter the fact that <the suspected beard> occurred on or below the chin GIVEN the ENACTMENT: "In this Act, beard means any facial hair no shorter than 5 millimetres in length that: occurs on or below the chin…" (/test/acts/47/4 1935-04-01)
print(beard_rules[1])
the Rule that the court MAY ALWAYS impose the RESULT: the fact that <the suspected beard> was a beard GIVEN: the fact that <the suspected beard> was facial hair the fact that the length of <the suspected beard> was at least 5 millimeter the fact that <the suspected beard> existed in an uninterrupted line from the front of one ear to the front of the other ear below the nose GIVEN the ENACTMENTS: "In this Act, beard means any facial hair no shorter than 5 millimetres in length that:…exists in an uninterrupted line from the front of one ear to the front of the other ear below the nose." (/test/acts/47/4 1935-04-01) "exists in an uninterrupted line from the front of one ear to the front of the other ear below the nose." (/test/acts/47/4/b 1935-04-01)
The difference between these two Rules is that the first one applies to facial hair "on or below the chin" and the second applies to facial hair "in an uninterrupted line from the front of one ear to the front of the other ear below the nose". I'll rename them accordingly.
chin_rule = beard_rules[0]
ear_rule = beard_rules[1]
AuthoritySpoke doesn't yet have a feature that directly takes a set of known Facts, applies a Rule to them, and then infers legal conclusions. Instead, in its current iteration, AuthoritySpoke can be used to combine Rules together to make more Rules, or to check whether Rules imply or contradict one another.
For instance, if we create a new Rule that's identical to the first Rule in the Beard Tax Act except that it applies to facial hair that's exactly 8 millimeters long instead of "no shorter than 5 millimetres", we can determine that the original "chin rule" implies our new Rule.
beard_dictionary[0]['inputs'][1]['content'] = 'the length of the suspected beard was = 8 millimetres'
longer_hair_rule = readers.read_rule(beard_dictionary[0], client=legis_client)
print(longer_hair_rule)
the Rule that the court MAY ALWAYS impose the RESULT: the fact that <the suspected beard> was a beard GIVEN: the fact that <the suspected beard> was facial hair the fact that the length of <the suspected beard> was exactly equal to 8 millimeter the fact that <the suspected beard> occurred on or below the chin GIVEN the ENACTMENT: "In this Act, beard means any facial hair no shorter than 5 millimetres in length that: occurs on or below the chin…" (/test/acts/47/4 1935-04-01)
chin_rule.implies(longer_hair_rule)
True
Similarly, we can create a new Rule that says facial hair is never a beard if its length is greater than 12 inches (we'll use inches instead of millimeters this time, and the units will be converted automatically thanks to the pint library). And we can show that this new Rule contradicts a Rule that came from the Beard Tax Act.
beard_dictionary[1]["despite"] = [
beard_dictionary[1]["inputs"][0],
beard_dictionary[1]["inputs"][2],
]
beard_dictionary[1]["inputs"] = {
"type": "fact",
"content": "the length of the suspected beard was >= 12 inches",
}
beard_dictionary[1]["outputs"][0]["truth"] = False
beard_dictionary[1]["mandatory"] = True
long_means_not_beard = readers.read_rule(beard_dictionary[1], client=legis_client)
print(long_means_not_beard)
the Rule that the court MUST ALWAYS impose the RESULT: the fact it was false that <the suspected beard> was a beard GIVEN: the fact that the length of <the suspected beard> was at least 12 inch DESPITE: the fact that <the suspected beard> was facial hair the fact that <the suspected beard> existed in an uninterrupted line from the front of one ear to the front of the other ear below the nose GIVEN the ENACTMENTS: "In this Act, beard means any facial hair no shorter than 5 millimetres in length that:…exists in an uninterrupted line from the front of one ear to the front of the other ear below the nose." (/test/acts/47/4 1935-04-01) "exists in an uninterrupted line from the front of one ear to the front of the other ear below the nose." (/test/acts/47/4/b 1935-04-01)
long_means_not_beard.contradicts(ear_rule)
True
Finally, let's look at adding Rules. AuthoritySpoke currently only allows Rules to be added if applying the first Rule would supply you with all the input Factor you need to apply the second Rule as well. Here's an example.
The Beard Tax Act defines the offense of "improper transfer of beardcoin". This offense basically has three elements:
But in section 7A of the Beard Tax Act, we also learn specifically that a "loan" of the tokens called beardcoin counts as the kind of "transfer" that will support a conviction of the offense. We can represent this information as a separate Rule, and then add it to the Rule defining the offense. The result is that we discover an alternate way of establishing the offense:
Here are the two Rules we'll be adding together.
elements_of_offense = beard_rules[11]
print(elements_of_offense)
the Rule that the court MUST ALWAYS impose the RESULT: the fact that <the defendant> committed the offense of improper transfer of beardcoin GIVEN: the fact that <the beardcoin transaction> was a transfer of beardcoin between <the defendant> and <the counterparty> absence of the fact that <the beardcoin transaction> was a licensed beardcoin repurchase the fact it was false that <the counterparty> was <the Department of Beards> DESPITE: the fact that the token attributed to <the Department of Beards>, asserting the fact that <the Department of Beards> granted an exemption from the prohibition of wearing beards, was counterfeit GIVEN the ENACTMENTS: "It shall be an offence to buy, sell, lend, lease, gift, transfer or receive in any way a beardcoin from any person or body other than the Department of Beards, except as provided in Part 4." (/test/acts/47/7A 1935-04-01) "It shall be no defense to a charge under section 7A that the purchase, sale, lease, gift, transfer or receipt was of counterfeit beardcoin rather than genuine beardcoin." (/test/acts/47/7B/2 1935-04-01) DESPITE the ENACTMENT: "The Department of Beards may issue licenses to such barbers, hairdressers, or other male grooming professionals as they see fit to purchase a beardcoin from a customer whose beard they have removed, and to resell those beardcoins to the Department of Beards." (/test/acts/47/11 2013-07-18)
loan_is_transfer = beard_rules[7]
print(loan_is_transfer)
the Rule that the court MUST ALWAYS impose the RESULT: the fact that <the beardcoin transaction> was a transfer of beardcoin between <the defendant> and <the counterparty> GIVEN: the fact that <the beardcoin transaction> was <the defendant>'s loan of the token attributed to <the Department of Beards>, asserting the fact that <the Department of Beards> granted an exemption from the prohibition of wearing beards, to <the counterparty> GIVEN the ENACTMENT: "It shall be an offence to buy, sell, lend, lease, gift, transfer or receive in any way a beardcoin from any person or body other than the Department of Beards, except as provided in Part 4." (/test/acts/47/7A 1935-04-01)
But there's a problem. The loan_is_transfer
Rule establishes only one of the elements of the offense. In order to create a Rule that we can add to elements_of_offense
, we'll need to add Facts establishing the two elements other than the "transfer" element. We'll also need to add one of the Enactments that the elements_of_offense
Rule relies upon.
loan_without_exceptions = (
loan_is_transfer
+ elements_of_offense.inputs[1]
+ elements_of_offense.inputs[2]
+ elements_of_offense.enactments[1]
)
print(loan_without_exceptions)
the Rule that the court MUST ALWAYS impose the RESULT: the fact that <the beardcoin transaction> was a transfer of beardcoin between <the defendant> and <the counterparty> GIVEN: the fact that <the beardcoin transaction> was <the defendant>'s loan of the token attributed to <the Department of Beards>, asserting the fact that <the Department of Beards> granted an exemption from the prohibition of wearing beards, to <the counterparty> absence of the fact that <the beardcoin transaction> was a licensed beardcoin repurchase the fact it was false that <the counterparty> was <the Department of Beards> GIVEN the ENACTMENTS: "It shall be no defense to a charge under section 7A that the purchase, sale, lease, gift, transfer or receipt was of counterfeit beardcoin rather than genuine beardcoin." (/test/acts/47/7B/2 1935-04-01) "It shall be an offence to buy, sell, lend, lease, gift, transfer or receive in any way a beardcoin from any person or body other than the Department of Beards, except as provided in Part 4." (/test/acts/47/7A 1935-04-01)
With these changes, we can add together two Rules to get a new one.
loan_establishes_offense = loan_without_exceptions + elements_of_offense
print(loan_establishes_offense)
the Rule that the court MUST ALWAYS impose the RESULT: the fact that <the beardcoin transaction> was a transfer of beardcoin between <the defendant> and <the counterparty> the fact that <the defendant> committed the offense of improper transfer of beardcoin GIVEN: the fact that <the beardcoin transaction> was <the defendant>'s loan of the token attributed to <the Department of Beards>, asserting the fact that <the Department of Beards> granted an exemption from the prohibition of wearing beards, to <the counterparty> absence of the fact that <the beardcoin transaction> was a licensed beardcoin repurchase the fact it was false that <the counterparty> was <the Department of Beards> GIVEN the ENACTMENTS: "It shall be no defense to a charge under section 7A that the purchase, sale, lease, gift, transfer or receipt was of counterfeit beardcoin rather than genuine beardcoin." (/test/acts/47/7B/2 1935-04-01) "It shall be an offence to buy, sell, lend, lease, gift, transfer or receive in any way a beardcoin from any person or body other than the Department of Beards, except as provided in Part 4." (/test/acts/47/7A 1935-04-01)
There will be additional methods for combining Rules in future versions of AuthoritySpoke.
For now, try browsing through the beard_rules object to see how some of the other provisions have been formalized. In all, there are 14 Rules in the dataset.
len(beard_rules)
14
The Beard Tax Act example still presents challenges that AuthoritySpoke hasn't yet met. Two capabilities that should be coming to AuthoritySpoke fairly soon are the ability to model remedies like the sentencing provisions in /test/acts/47/9, and commencement dates like the one in /test/acts/47/2.
But consider how you would model these more challenging details:
The "purpose" provisions in /test/acts/47/3 and /test/acts/47/10
Provisions delegating regulatory power, like /test/acts/47/6B and /test/acts/47/12
Provisions delegating permission to take administrative actions, like /test/acts/47/6/1
Provisions delegating administrative responsibilities, like /test/acts/47/6D/1 and /test/acts/47/8/1
Provisions delegating fact-finding power, like /test/acts/47/6D/2
Clauses limiting the effect of particular provisions to a certain statutory scope, like the words "In this Act," in /test/acts/47/4
For more about the use of the Beard Tax Act to describe the effectiveness of legal data modeling software, see the Python for Law Blog.
If you have questions, comments, or ideas, please feel welcome to get in touch via Twitter at @AuthoritySpoke or @mcareyaus, or via the AuthoritySpoke Github repo.