Should We Make a Guild for AI Work?

I recently reread How professional Ethics work by Sibylla Bostoniensis and it resonated with other ideas about data science titles and care about the craft. So here goes my rambling:

Should we make a guild for AI work?

for love of the craft

an image of assorted wood working tools on a wall

Every once in a while, fellow data scientists complain about other data scientists. Some gate-keeping comments like: “These youngsters don’t know anything about data science work in practice”, “Anyone who participated in one Kaggle competition calls themselves a data scientist nowadays”, “People who were data analyst, are calling themselves data scientists now”, “That is not a data scientist, that is a database administrator who took one python course”, “That is not a data scientist, that is a python programmer who just downloaded a model and runs it without thinking”.

Some people complain because their job used to be exclusive and highly paid, and the job market is now muddied. They want exclusivity above all, and want to put themselves in the exclusive group, and everyone else out. I have a slightly different view, I don’t want to keep people out. I want to allow as much people in. As much variety as possible, an inclusive place for everyone. That is why I write and try to level up my colleagues.

Some of the critiques are valid. I do care about the craft of machine learning, and coding in general. Applying machine learning models without thought is problematic. Creating automated decision-making at scale can become problematic very quickly (you don’t even need machine learning to make it horrible)! And so it could be beneficial if people who create and approve automated decision-making are held accountable. Hospitals and doctors are held accountable, CEO’s used to be held accountable for bad decision-making, so why not data scientists?

It is in this mix of accountability and love and care for the craft that we find the idea of an AI guild as solution.

The guild could set standards, keep people up to date on the newest techniques, set ethical and professional guidelines for when to apply techniques and when not. And as an external organization, it can externally validate someone’s claim of professionalism.

The idea of a guild

The idea about craftsmanship is not new, there are many books about craftsmanship in coding. The pragmatic programmer: from Journeyman to master (1999) loved this idea. And it has great potential, official training and official recognition could work out in our favor.

How do guilds work?

(from wiki) A guild is an association of artisans who oversee the practice of a craft in an area. Somewhere between a professional association, a trade union, a cartel and a secret society. While there were professional groups in Roman times, they were voluntary. In Europe in the High Middle Ages, guilds arose, as craftsmen united to protect their common interests. Guilds are made up of experienced and confirmed experts in the field.

Guilds have the privilege to sell their goods or practice their skill within an area. Only guild members are allowed to work. The guild sets hourly prices and determines what quality is good enough. The guild also determines standing in the guild. Members rise in the ranks from apprentice to journeyman and finally master or even grand master.

The idea of gradual progression in the data science world and clearly marking a ranking of someone sounds appealing to people. Instead of people just calling themselves data scientist and you as a company having to dig and dig to figure out if someone can do the job that you want to be done, we have an AI guild. The guild will tell you what rank someone has. The AI guild will only make someone journeyman if they take an apprenticeship where they learn the finer points of AI work. After many years of journeyman, someone has to deliver a master proof and becomes an AI master, recognized by the entire guild and thus the entire world.

There is an actual AI guild online, but it is voluntary, and also it feels more like a company than a proper guild. Making a guild voluntary makes it too easy to just choose someone who is not a member.

From guild to exclusive professional organization

To make a guild work, we should probably follow the examples of professional organizations like doctors. That means that the government determines that the only people allowed to work on AI are people who are part of the professional organization. Only licensed AI professionals and no one else. The professional organization has professional ethics; binding rules for what a member can, and cannot, do. A board that will enforce the rules and can kill your career by withdrawing your license. It would mean you are bound to more rules than a normal person, but you get the exclusive right to perform AI tasks.

There are a lot of people with libertarian ideas in the programming world, and binding yourself to a professional organization that has authority over you is a hard sell. But I think that is the only way to get your (exclusive) guild. Voluntary groups have to fight for funding and recognition. We know tech giants hate unions, and so they will probably work around voluntary organizations. It is very easy to find cheaper workers somewhere else. Or to set up a competing organization that only cares about code quality and none of those messy ethics.

There are some hopeful signs, with the new AI act in Europe, with stringent rules about applying AI in governments, it might be beneficial if you hire people who have bounded and enforceable ethical rules. The guild enforces the rules and so by hiring those people, you automatically make it easier to comply to regulations because all guild members share the same code of conduct.

Will it work?

hand of a person using a chisel while carving wood

Licensed professionals in this world are lawyers, physicians, psychotherapists and some other medical health professionals. There are also architects and engineers in professional organizations. But not in programming and not in data science. Could we do it? Are we able to set expectations about what constitutes a journeyman level of Machine learning? Who do we recognize as masters? What would the professional ethics of such a group look like? How can we create a diverse group that represents all data scientist’s interests and enforces standards?

The difference between doctors and data scientists is that data science work is (mostly) non-dangerous. Doctors do work that is normally forbidden, but an exclusive right has been given to people of the doctor profession. Practicing law is not allowed without a license. Data science work is undefined, what does it constitute? Am I doing data science work if I use a pre-baked model? Or when I train it? Is a statistician a data scientist?

I’m sorry, I don’t think there is an easy answer, I have only produced more questions. Ultimately, I think there might be a place for professional organizations in the machine learning world, but probably only in very restricted areas. Since everyone can do AI and no one is held responsible for the work that their automation (AI) does, I don’t think a guild will be any help.