Serving Algorithms

Nearly three decades of research and development has resulted in nine core algorithms that enable Smartsettle’s intelligent solutions. Smartsettle’s neutral site is in the business of serving these algorithms to multiple decision makers with conflicting objectives in order to help promote agreement.

The neutral site, through a special API, receives information from its users and responds by serving information that helps collaborative negotiators reach fair, efficient, and timely agreements. In short this is a “collaborative negotiation service”.

The Smartsettle algorithms are implemented at the neutral site in order to support the Smartsettle Software As A Service (SaaS) model. They typically operate in the following order:

1. Single Negotiating Framework℠ (SNF)

establishes the context for the negotiations and any conditions that will apply once agreement has been reached

2. Comprehensive Preference Analysis℠ (CPA)

enables the system to learn the underlying priorities and interests of the parties

3. Visual Blind Bidding℠ (VBB)

lets parties exchange visible proposals and Suggestions and secretly accept any they could sign, expediting the path to agreement

4. Share The Overlap℠ (STO)

declares agreement as soon as the parties' secret acceptances overlap, so no further bargaining is needed

5. Reward Collaborative Behaviour℠ (RCB)*

determines where within an overlap the deal falls, and how a small gap is divided, by favouring the party who made the smallest last move, which discourages holding back and motivates true collaboration

6. Small Gap Closer℠ (SGC)**

closes a small remaining gap using RCB, more fairly than the traditional "split the difference"

7. Expert Neutral Decider℠ (END)

offers a voluntary last resort: when both parties choose it, the Expert Neutrals' opinions of fair, combined with RCB, produce a guaranteed settled value

8. Fairness Enhancing Normalization℠ (FEN)

normalizes satisfaction scales so that MMG treats all parties fairly and cannot be gamed

9. Maximize the Minimum Gain℠ (MMG)

always caps the process, harvesting any hidden value that remains after a Baseline and distributing it fairly among all parties

* also known as Smallest Last Move (SLM)

** formerly called the Automatic Deal-Closer (ADC)

The process starts with all parties agreeing to collaborate in building an SNF, which is like the final agreement except for blanks representing unresolved issues. After the framework is accepted by all parties, complete with negotiating ranges for all the issues, the next step is to model the problem within the Smartsettle interface. Through a process of preference elicitation, CPA determines how parties become satisfied on each issue. Smartsettle is then able to help them with the creation, rating, and exchange of proposals and secret acceptances on generated suggestions through VBB.

VBB can reach agreement in four ways, in increasing order of severity. A party may simply accept a visible proposal. More commonly, the parties' secret acceptances overlap: STO declares the deal and RCB determines where it falls. If the acceptances come close without overlapping, SGC closes the small gap using RCB. If Final Session ends with the parties further apart, parties who have voluntarily opted in to the END receive a settled value based on the opinions of Expert Neutrals, again with RCB favouring the party closest to fair. In multivariate cases any of these agreements becomes a Baseline, and MMG, made fair by FEN, then uncovers any hidden value that remains. For a fuller explanation, see Rewarding Good Negotiating Behaviour smartsettle.com.

Behind the scenes, RCB makes it counterproductive to hold back. Resulting increased settlement rates save time and money in all types of negotiations and promote access to justice in dispute resolution. Impasse is avoided when there’s a small gap and the SGC uses RCB to close it instantly and without bias. If all else fails, an agreement can be guaranteed if parties agree to Smartsettle’s Expert Neutral Decider (END), which will guarantee a collaborative outcome.

Normalization and improvement algorithms are relevant only to multivariate problems. Normalization with FEN encourages parties to focus on what they want by making sure that no party is able to game the system. The final step in multivariate cases is MMG to make sure that no value is left on the table. If parties have their preferences well-represented, Smartsettle can generate and fairly distribute mutually beneficial gains.

The Smartsettle service is used in simulated negotiations, games, and just starting to be used in real-life. In February, 2019, Smartsettle ONE made history as the first time AI was used for dispute resolution in a public court in England and Wales. According to the Law Gazette, after a certain dispute over a £2,000 unpaid bill for a personal counselling course had being making little progress over the previous three months. Smartsettle ONE was able to help the parties find a satisfactory outcome in the space of an hour.

Parties may be trained to use the system themselves or engage one or more skilled facilitators. A human mediator is optional because the Smartsettle system itself provides mediation functions. If you have a simple negotiation that could benefit from Smartsettle’s service, you may wish to first take Smartsettle ONE for a test drive by practicing against the Smartsettle ONE robot at our demonstration site. If your case can easily be reduced to a single numerical issue you can invite your opponent to also practice before you negotiate for real. Facilitators and mediators are available if you need a hand. Request Smartsettle Infinity if your case is more complex.

If you already have your own application and would like to enhance it with Smartsettle powered intelligence this can be done quickly and painlessly with a branded site that subscribes to our neutral site via an API. Branded sites can be configured to suit a wide variety of needs. If you have a special need that is not represented by our configuration options then talk to us about a plugin.