Looking at AI (Artificial Intelligence)Use in Mediation

It's pretty much impossible in the modern world to avoid AI. It is everywhere - event after event, article after article and advert after advert. Within the mediation space, where I work as a civil,commercial and employment mediator, I have seen people hail it as the end ofmediation; its next destination or just an organic part of the process.Whatever the topic, AI is presented as the future - may we all be grateful tolive in such an age.

And yet, on the ground in actualmediations, AI use is frequently problematic. It can and does cause real-worldproblems for parties. It can enable parties to adopt unrealistic positions thatwould be unsustainable once they hit a courtroom; it can create"solutions" that don't actually work for anyone; and overall itcan cause parties to understand and care less about the other party and to be more,not less distanced. These are all problems that the mediatorworking with parties using AI needs to be able to handle right now.

This article is not intended to be a polemic against AI. I'm not trying to be a luddite; nor am I trying to claim to be an expert on AI. There are good uses for AI and there is potential for its use within mediation. The issue comes with how parties use AI and themistakes that the human beings using it make. These are mistakes that a numberof mediators that I work with are starting to see crop up in cases. In thisarticle, what I want to do is explore 5 of the biggest issues with AI foron-the-ground mediation, and how parties and mediators can work with to avoidthese.

Using AI for Mediation Statements andPositions

One of the biggest dangers with AI is inhow parties develop their own positions in advance of the mediation. Wehave all seen the capability of AI to produce vast swathes of text, which looksgreat at first glance. This seemingly helps parties to create their argumentand find their voice, without having to put the hours in of actually writingthe thing.

The problem with these positions are: it'snot their voice or message; its AI's. And this lacks authenticity. Italso lacks ownership of the concepts expressed within it. Authenticityis a cornerstone to mediation. The parties need to trust and believe that theother party actually thinks what it is saying.

As a mediator, I feel I can work withalmost any viewpoint or need if it is genuinely held. You think that the sky isred - no problem. If you genuinely think that, let's explore your perspectiveand reasoning and how that affects how you see the case and what you need. Buta viewpoint that you don't actually have (or worse, have not actuallyunderstood or realised is in your AI generated bundle), this is a nightmare towork with. It cannot be explored further - you either don't hold the belief orit's not that important to you. If the other party make concessions based onit, you're not getting what you need. And if the other party starts to perceiveyou as inauthentic in the mediation, you are absolutely sunk. They will ignoreall of your requests as posturing.

Mediation works best when parties are ableto say what they actually think, mean and need. This allows the other party towork with them to find solutions. AI should be used to support this trueposition, not hinder or hide it.

Therefore, Rule 1: Be authentic. If you are going to use AI, use it to enhance what you actually think and want. Not to auto-generate an argument that you don't have.

AI to help the Reader, not the Writer

The second big mistake with AI use inpositions is that it is commonly written entirely from the point of view ofthe party creating it. Little thought (if any) is given to how it's goingto be read by the other party. This is a huge problem. When we think about mostwriting, it is created with the idea of a reader in mind. We write a letter tosomeone - and we would change our style and message based on whether thatperson was a prospective employer, an ex-spouse, a child, the UN SecretaryGeneral, etc and what their interests are.

Frequently with AI, people generate textto solve a task. I want to have written aletter to the UN Secretary General, not I want to write a letter to the UN Secretary General. The reader of the AI text is the person who generated it and they are checking it to see if it satisfies the task. They are not checking whether it is actually right for the end reader, but whether it does what they needed to do so they can get it off their desk. The failure to read AI-generated text critically is exacerbated by the effects of confirmation bias- we check to see it's matching up with what we want it to be, not reading it-and the Barnum Effect - this is the psychological effect people experienc where they misjudge a generic vague text as being specific to their situation, such as with a horoscope. This all leads to text that is seemingly good enoughfor the user but lacks depth and precision.

The problem with all of this is that the end-reader is forgotten about. When they do receive it, they may not find it at all relevant to them. AI-generated text also often lacks the cadence and human voice that makes an argument compelling or persuasive. This all just means that the text becomes hard to read and unengaging. It also will often have none of the points that are important to the other party because the AI-user is unaware of them or hasn't thought to put them in. Within a mediation, positions canseem more positional and less-engaging when AI is overused.

In generating text, you always need to think about who are you are writing to and adapt. Many people I know who use AI will draft multiple times and fine-tune. What should be avoided is doing it just to have got it done.

So, Rule 2: Use AI to engage with the other party, not to solve a task you need to get done.

Be careful on AI for advice

The third biggest issue follows on from theabove, where people use AI for advice about things that they themselvesdon't know. This is an extremely dangerous territory to be in as the party has no idea if the points generated are correct or not. The party who uses AI in this way - especially with legal or technical advice - frequently gets themselves into problems when faced with an actual expert on the matter. What makes this worse, is that parties often don't know what they don't know and so either become completely unstuck here or doggedly insist on going on, trusting the AI to the end. Using AI as your lawyer, is like using WebMD as your brain surgeon, it's ill advised and no-one should be surprised when you end up with a splitting headache.

This is again not to say that you couldn'tuse AI, but you need to be able to check the information. If you are unable to check the information, you shouldn't put it in or you should get anactual expert adviser to check it (or produce it to start with). Additionally, when doing the research parties need to think widely about the task that they are doing and from a range of perspectives. Rather than starting with "generate a legal argument for all the reasons to support my position", it would be useful to also run "generate all of thereasons that support the other side and why I should settle" but sadly too few people, if anyone does this.

Rule 3: Only use AI to generate contentthat you have the expertise to be able to check for accuracy.

AI for Option Generation

With option generation, I've seen AIused in a few ways. The best way to use AI if you are going to do this, iscollaboratively with the other side. That way the parties are actuallyworking together on something. This is positive for the likelihood of the solutionactually being practicable and what the parties want and it also reflects theiterative process that mediation works best with.

Where option generation goes awry with AI,is when parties use either from their own position or just on complete blue-skythinking mode. This is when people create wacky ideas which they don'tappreciate are insults to the other party, or lack credibility if passed. Togive an example, I can think of a case where AI was used to generate sixoptions of which one of the options relied on something that orally had beenpassed as something the other side had no interest in (and with goodreasoning). Therefore, exploring that options would have exacerbated tensionsas it would have seemed like the side who were using AI did not get the otherparty or understand what mattered. It is also reduces your own credit from theother party as the problem-solving party, if it appears that you are coming upwith solutions only through auto-generation, rather than through actuallylistening and understanding the other party and applying thought. Why would yougive this valuable benefit away?

So, as a fourth rule: Be explicit ifusing AI to generate options and also use it alongside human thought andunderstanding. It works best when used collaboratively with the other partyand don't forget the power of the iterative journey in working with others tofind a solution.

AI and Human Relationships in Mediation

Possibly the biggest issue with overusingAI in mediation is a fundamental one in that parties who overuse it (or use it badly), neglect the importance of the human relationship in mediation.

Mediation is a human practice. As a mediator, I am always working with people - whether it's a mediation between two individuals or two multi-nationals. We want to work with people - we want them to understand us; we want to understand them. We remember how people made us feel - good and bad. We want to like, love and work with other people. We also want to forgive them, build bridges with them and come to agreements with them - even if that agreement means going a separate way.

Parties and advisers need to be reallycareful about creating barriers to this human need. Whilst there are stories of people falling in love with AI machines, the average person is only going to say yes to a marriage proposal from a person they love; only receivean apology well from a human being they can see emphasises with them; and only sign a settlement agreement with someone they trust and who they think will honour the agreement.

This needs humanity, human interactionand human respect. It needs people to realize that it's not just what you say, but how you say it and whether you mean it that matters. People like people because we are all people. It is what unites us and allows us to find ways through conflict. We understand others' viewpoints and needs because we have the capacity to engage with it and it is this capacity which gives us the potential to find a resolution which satisfies two different human experiences and needs.

So, Rule 5: Never underestimate theimportance of being human. People like people and being human in a conflictsituation should come first. AI should not take precedence to humanity if youwant to get a human being to agree.

Conclusion

As stated at the start, the aim of thisarticle is not to say don't use AI in Mediation. But it does need to be usedwith caution and care by parties and I don't believe it is the panacea thatsome have made it out to be.

Mediation, in my opinion, will always be ahuman experience and the more complicated or sensitive the case, the moreimportant it is to remember the humanity.

To summarise, I think that there are fivekey rules for using AI in mediation.

1. Be authentic - do not let AI talk     for you but enhance your own voice. Own the content generated.
2. Use AI to engage with the other party, not just to talk from your own perspective
3. Be very careful on using AI for expertise that you do not have. Never use it to generate something you are unable to check     for accuracy.
4. Use AI for option generation with the other party and remember the importance of working iteratively.
5. Never understimate the importance of being Human. People connect with people. Prioritise human-connection.

‍