Reading: Difference between revisions

429 bytes added ,  2 February 2020
no edit summary
No edit summary
No edit summary
Tags: Mobile edit Mobile web edit
Line 5: Line 5:
Before replacing your lawyers with [[artificial intelligence]], consider the [[lawyer]]’s basic function: ''[[reading]]''.  However, this is all any contemporary artificial intelligence can do. Even a deep neural network.
Before replacing your lawyers with [[artificial intelligence]], consider the [[lawyer]]’s basic function: ''[[reading]]''.  However, this is all any contemporary artificial intelligence can do. Even a deep neural network.


Consider what “junior legal work” tends to comprise. Reading and interpretation of basic contracts: the paradigm case is the [[confidentiality agreement]]. It is a [[tedious]] low-value job. Just when sales are on the brink of a grand new conquest, the prospective client produces a confidentiality agreement. Sales is ready to go, but first thee [[legal eagle]]s must have their day. Cue a week or so of tedious, costly, back-and-forth, during which time ardours run a big risk of being dampened.
Consider what “junior legal work” tends to comprise. Reading and interpretation of basic contracts: the paradigm case is the [[confidentiality agreement]]. It is a [[tedious]] low-value job. Just when sales are on the brink of a grand new conquest, the prospective client produces a confidentiality agreement. Sales is ready to go, but first thee [[legal eagle]]s must have their day. Cue a week or so of tedious, costly, back-and-forth, during which time ardours are in grave danger of dampening.


What to do? {{t|Policy}} says we must review the {{t|contract}}. But surely, in these [[AI|artificially intelligent]] times, there is a better way. Could not a hastily-commissioned [[chatbot]] handle this?
What to do? {{t|Policy}} says we must review the {{t|contract}}. But surely, in these [[AI|artificially intelligent]] times, there is a better way. Could not a hastily-commissioned [[chatbot]] handle this?


Good luck on that soldier. First clue here is that not even the salesperson — supposedly a reflective, emotionally aware, tertiary-educated, autonomous intelligent being<ref>''[[Salesperson|Allegedly]]''.</ref> — can be trusted to review this contract. If she could, you wouldn’t need the [[legal eagle]], would you?  
Good luck on that, soldier. First clue here is that not even the [[salesperson]] — supposedly a reflective, emotionally aware, tertiary-educated,<ref>Granted, in something useless, like ancient Greek.</ref>1 autonomous intelligent being<ref>''[[Salesperson|Allegedly]]''.</ref> — can be trusted to review this contract. If she could, you wouldn’t need the [[legal eagle]], would you?  


So there is your first hurdle: your [[chatbot]] may not need the forensic skills of Judge Learned Hand himself, therefore, but it must at least be better at reading a {{t|contract}} than a [[salesperson]].  
So there is your first hurdle: your [[chatbot]] may not need the forensic skills of [[Cardozo J]] himself, therefore, but it must at least be better at reading a {{t|contract}} than a [[salesperson]].  


And what must that review entail? Reading is more than the mechanical ingestion and processing of a string of symbols. When a lawyer reads a contract she is doing something that even a university-educated [[salesperson]] cannot. Reading and interpretation is a dynamic process by which the reader brings her personal metaphorical superstructure — a “schema” — to a text that was composed using a more or less compatible schema.  No two schemas are the same — we all have our foibles and unique experiences, and those variances in everyday life account for much of the human condition. Lawyers have their own special meta-schema — one that requires years learning and refinement.  The lawyer uses this meta-schema to extract meaning and consequences that are unavailable to laypersons. Legal [[magic words]] have special meanings: “[[indemnity]]”; “[[consideration]]”; “[[equitable remedy]]”.  Concepts like these have their own intellectual life and a dog-eared, meandering history which one can trace through centuries of dusty law reports. When she reads a contract, a lawyer brings her own imperfect<ref>And it will be imperfect: most commercial lawyers, for example, have a very dim grip on the concept of an [[indemnity]] for example.?</ref>, idiosyncratic impression of that history to her review.
And what must that review entail? Reading is more than the mechanical ingestion and processing of a string of symbols.<ref>Indeed, calling computer code a string of “symbols” is itself a profoundly misleading metaphor. Symbols are symbolic of something. They call to mind — “mind” — an alternative intellectual conveyor and invite the nine to compare them, and draw an imaginative analogy. Nothing remotely like that happens when a machine processes a string of code.</ref> When a lawyer reads a contract she is doing something that even a university-educated [[salesperson]] cannot. Reading and interpretation is a dynamic process by which the reader brings her personal metaphorical superstructure — a “schema” — to a text that was composed using a more or less compatible schema.  No two schemas are the same — we all have our foibles and unique experiences, and those variances in everyday life account for much of the human condition. Lawyers have their own special meta-schema — one that requires years learning and refinement.  The lawyer uses this meta-schema to extract meaning and consequences that are unavailable to laypersons. Legal [[magic words]] have special meanings: “[[indemnity]]”; “[[consideration]]”; “[[equitable remedy]]”.  Concepts like these have their own intellectual life and a dog-eared, meandering history which one can trace through centuries of dusty law reports. When she reads a contract, a lawyer brings her own imperfect<ref>And it will be imperfect: most commercial lawyers, for example, have a very dim grip on the concept of an [[indemnity]] for example.?</ref>, idiosyncratic impression of that history to her review.


A [[neural network]] can have none of this. Nor can it acquire any of it through ingestion of sample texts.
A [[neural network]] can have none of this. Nor can it acquire any of it through ingestion of sample texts.