As a doctoral student at Cornell University, Ernest Thiessen used game theory to develop an efficient methodology for complex negotiations known as Interactive Computer-Assisted Negotiation Support System or ICANS, the first secure multiparty negotiation tool of its kind, known by many now as SmartSettle ONE® and SmartSettle Infinity®.

With SmartSettle ONE®, negotiators communicate via a secure neutral server on the Internet that acts as a mediator using AI algorithms to suggest efficient outcomes. The negotiators may interact entirely online or some combination of face-to-face, if they so choose.

Smartsettle Infinity® is able to handle multiple-party and any number of quantitative or qualitative issues with collaborative decision making. It’s perhaps the most advanced negotiation system available. The process involves parties first submitting initial proposals and the program elicits the their preferences. The parties are encouraged to accurately present their positions as misrepresentation will lead to inferior outcomes.

SmartSettle provides negotiators with a graphic interface panel that displays their proposals as well as those it generates. When negotiating multiple issues, the proposals are presented as packages comprised of values for all the issues. Parties may adjust their positions or accept a proposal in confidence, which is referred to in SmartSettle as a bid. When the system detects an overlap of bidding an agreement is declared.

Negotiators may request an improvement to their agreement. SmartSettle then generates potential agreements based on the game theory algorithm “maximize the minimum gain.”

If negotiators desire to work collaboratively, and if they are willing to be forthright in expressing their preferences, the system will facilitate their negotiation. SmartSettle is being used successfully for separation agreements, environmental disputes, insurance claims and to model possible agreements for Brexit as well as for Ukraine and Russia.

One study has concluded that in some situations AI negotiating agents can outperform human beings in terms of deal optimization, but less able to deal with arbitrary human conversations, semantic issues, emotions, and participants not being forthright or negotiating collaboratively.

There are now many artificial intelligence programs for negotiation and mediation. Are there lessons mediators can learn from AI?

Here’s a look at one program:

Adjusted Winner® uses Decision Theory and Game Theory to distribute items or issues to disputants based on the values assigned to them by the parties involved. The process requires that the items or issues be divisible and described in numbers, such as dollars or percentages. Each disputant assigns a value to each of the items or issues in dispute for a total of 100 points. The Adjusted-Winner procedure determines a fair outcome based on the premise that items or issues go to whomever values them more. In the end both parties end up with the same number of points. 

For example: If two siblings are disputing their parent’s estate they might assign values as shown below: 

Item                                                 Sally             Jane

Checking account                        50                 40

Home                                         20                 30

Cabin                                         15               1

Investments                                  10                 1

Other                                           5                  10

Total                                                  100                100

Adjusted Winner works by initially assigning, the item to the person who puts more points on it. Thus, Jane gets the home, because she placed 30 points on it compared to Sally’s 20. Likewise, Jane also gets the items in the other category, whereas Sally gets the checking account and the cabin. Leaving aside the tied item (investments), Sally has a total of 65 (50 + 15) of her points, and Jane a total of 40 (30 + 10) of her points. This completes the “winner” phase of Adjusted Winner.

Because Jane trails Sally in points (40 compared to 65) in this phase, initially the investments on which they tie are allocated to Jane, which brings her up to 50 points (30 + 10 + 10). 

Adjusted Winner then determines that Jane receive the home (30 points), the other items (10 points), and the investments (10 points). Together with 1/6 of the checking account, Jane’s point total is now 56.67. Sally would receive the cabin (15 points). Together with 5/6 of the checking account, Sally’s point total is now 56.67. Thus, each person receives exactly the same number of points, as she values their allocations.

The key to fairness in Adjusted Winner, of course, is that the participants are honest in assigning their preferences for the items being negotiated, in other words, a negotiator could attempt to “game” Adjusted Winner by misrepresenting their preferences.

What can a human mediator learn from Adjusted Winner? During mediation disputing parties may argue over items that they neglect to disclose are important to them. Adjusted Winner suggests that encouraging parties to discuss the importantance of each item, helps to facilitate a resolution.