AI in Mediation - the good, the bad and the ugly

 
I like information. I love a good deep dive into a topic to help me be more informed about my options and how to move forward.   These days we have more resources than ever to facilitate that.  We’ve all used Dr. Google before reaching out to a trusted practitioner or used an AI resource to scout for better airfares or plan a trip.  
But these research options often come with red flags.  Dr. Google may claim you’re suffering from a deathly ailment, or the AI generated travel plan may highlight options that don’t exist in real life. I’ve experienced both!  The same is true for mediation. It is well-established that AI often hallucinates and lacks other critical elements for an ethical mediation.  (https://www.pon.harvard.edu/daily/mediation/ai-mediation-using-ai-to-help-mediate-disputes/)
I’ve been practicing mediation for 12 years and this year has brought the new challenge of AI into the process.  I’ve had repeated instances of clients quoting me “law” they learned on ChatGPT regarding retirement account division, alimony awards and child support – that have zero basis in actual Maryland law.  Or clients presenting full agreements or parenting plans written by AI – again, failing to take into consideration local rules, laws or the wishes of the other party.  While it is a shortcut, it truly is a “double-edged sword.” (https://www.kcba.org/?pg=News-Bar-Bulletin&blAction=showEntry&blogEntry=126105)  And while I always support parties safely talking and generating proposals/ideas outside the mediation process, if they are not using quality sources then they run the risk of derailing the process with false assumptions.  
And of course there is much more to divorce than facts and figures generated by a machine.  That’s the reason you hire a mediator or other qualified professionals.  The mediator will find the nuance in your situation, explore numerous, creative options, facilitate conversation, create empathy and a safe space for everyone to be heard and will assure that your agreement actually conforms to local law and process.  No machine can do that (not yet anyway!).  
AI is cautionary for the mediation process in several ways: 

• Absence of Emotional Intelligence: AI cannot pick up on body language, tone, or interpersonal dynamics, which are crucial for navigating emotional disputes. This is why we mediate face to face (or via Zoom).
• Data Security and Confidentiality: Using public AI tools can expose private settlement information. Mediation is a confidential process – that is a core element.  When you feed your data into AI, you are violating the confidentiality principle (and possibly exposing your personal information).
• "Hallucinations" and Inaccuracy: AI tools may confidently generate false, legally unsound, or inaccurate information, leading to poor agreements. I am seeing this more and more and when someone quotes me “legal AI” that is more time in our process that needs to be spent debunking myths.  That costs everyone time and money. 
• Bias and Discrimination: AI systems can replicate or magnify biases present in their training data, leading to unfair outcomes. AI is essentially an echo chamber that tells you what you want to hear.  Data out is only as good a data in – so you can’t take what AI says as gospel.  Additionally, if you leave something out of your AI search, those very important nuance items will not be part of any result you get.  It’s like going to your doctor or therapist and not sharing the whole story.  Your outcome will reflect that.  Mediation is a neutral process.  
• Lack of Legal Context: AI tools often lack an understanding of specific local laws, court procedures, and the specific context of a case, resulting in non-contextual advice. Divorce laws vary greatly from state to state.  Making sure your agreement conforms to local laws, rules and process is critical. 
• Accountability and Liability: If an AI makes a mistake, it is unclear who is liable—the developer or the user. In mediation we are creating your agreement together, so you are always in control of the outcome.  I write each agreement personally and do not use AI as part of my process.  
• Misunderstanding Nuance: AI cannot effectively facilitate negotiation when parties are stuck, as it operates within strict programming rather than navigating complex human interaction. This is the beauty of the mediation process at its heart   - we are not just spitting out options, we are considering the nuance and impact in a way no algorithm can.  

 
There are many ways in which the use of AI in the mediation process in unethical and violates many of the main tenets of the process.  (https://www.cardozojcr.com/cjcr-blog/ai-mediation-ethically-questionable)
 
In mediation we can build trust between parties, which is central to a successful resolution. AI can’t do that.  In addition to making agreements in mediation, we are also learning HOW to resolve disputes and conflicts independently.  If you rely on a machine, you are missing out on the learning process. 
 
Ok – so lots of concerns, but how about the positives? I am by no measure anti-technology and believe it betters our life in many ways.  For mediation, AI is terrific at sifting through lots of information quickly, which can be really helpful for analyzing financial data or creating budgets.  It is also helpful in scheduling and other administrative actions.   
 
Mediation is a process that can’t be shortchanged by a machine – and while it may help with some research it will not replace the human element for a quality, lasting outcome.