How AI could soon save consumers from signing up to unfair terms and conditions

Most of us buy stuff on the internet without reading the terms and conditions. We take it for granted that the clauses of these standardized agreements are non-negotiable, and expect that they are in our best interests.

However, very often this is not what it seems. From laptop makers To airlines To Buy now, pay later companiesThere are endless rows about whether the terms and conditions are fair or not.

It’s hard to get solid facts about the size of this problem. But of course there are plenty of unhappy buyers on consumer forums.

Many, for example, have experienced implementation difficulties their contractual rights or was inadvertently What did they agree to when they bought something? Finding free or affordable legal advice in such circumstances can be difficult, so many people probably give up.

There’s an argument that people should read the contract before clicking “I agree,” but most of us don’t have the time or ability to do so. However, soon a solution will come out. Instead of solving all the small print ourselves, we may soon be able to use artificial intelligence.

what’s already there

AI tools to analyze legal documents have existed in very basic form for some time. they can do Identify potential issues such as infringement of rights that the consumer may wish to investigate further. But you’ll have to copy and paste the words sentence by sentence because the AI ​​is limited in how much text it can handle, and they’re designed to be a guide to reading yourself rather than to remove the blurb entirely .

are more sophisticated AI tools Which solves the related problem of reading web policy documents. Instead of pasting in text, you upload the relevant URL. The important but narrow focus here is on how web providers use your data. This makes it easy to teach an AI model everything it needs to know—especially when heavily regulated area,

With terms and conditions, the challenge is their varied nature. Sellers are more free to frame everything in their own words, which makes it more difficult for AI to detect and understand.

There are also many variations between different jurisdictions, such as “solicitor” in the UK and “lawyer” in the US. This means that AI trained on US data could mislead UK consumers. Yet existing instruments are often unclear as to which jurisdiction they are designed for.

You may be wondering if the alternative may be to copy and paste the terms and conditions into one of the latest AI chatbots chatgpt, but that too is not a solution. These general model Are not specifically trained on legal texts or legal analysis. This means that any advice they give has the potential to be accurate, inaccurate or completely fabricated.

problem solving

As far as we know, no team of developers is trying to create a dedicated terms and conditions AI for consumers using Open AI’s GPT-4-like model, which underpins ChatGPT.

Instead, many AI developers are focusing on a more lucrative field. tool making Which will automate legal work for law firms and other companies. This may also lead to less favorable terms for consumers, as the focus will be on cutting costs rather than improving quality of service.

To change this situation, lead author Jens Krebs and his colleague Ella Haig at the University of Portsmouth have been developing A Terms and Conditions app for England and Wales. When fully developed, it will enable people to copy and paste entire documents into the prompt.

It will then list any terms that may affect the consumer unduly, for example, by failing to meet legislative standards. Consumer Rights Act 2015, It will also compare all the terms used by comparable sellers to ensure that nothing unusual has been left out. When it sees something unusual, it will advise the user to read that part before deciding how to proceed.

The project is currently in the phase of testing the app on different AI models to see which one is most effective. So far Google’s BERT is coming out best with 81% accuracy, testing it against data where researchers know what the correct result should be.

Nothing will be launched until accuracy hits 90% to 95%. It is expected that the app will be used by consumer groups such as the Who? Go to general release in 2024 and again in 2025. The intent is that it will be free to use.

The main stumbling block for such a project is the lack of examples of harmful terms to train the AI ​​on – exactly the same problem consumers face if they are brave enough to try to judge terms and conditions. The long-term plan to increase accuracy in the Portsmouth app is to supplement and replace its training data with real data from consumer organisations, government and consumers.

The hope is that the app will be at the forefront of a new generation of AI tools designed to make terms and conditions less opaque. As well as potentially reducing the number of unhappy consumers, they can also help those who have already signed unfair terms to prepare and present their case – reducing the need for lawyers It happens.

If such services are discontinued, it can be expected that they also discourage vendors from pushing the boundaries of what is acceptable. If the terms and conditions turn out to be a bit more consumer friendly, it would be a huge win for this emerging technology.

This article is republished from Conversation Under Creative Commons Licence. read the original article,