Automated bias elimination in negotiated terms

Patent metadata

Field	Value
Publication number	US-11907820-B2
Application number	US-201816203404-A
Country	US
Kind code	B2
Filing date	Nov 28, 2018
Priority date	Nov 28, 2018
Publication date	Feb 20, 2024
Grant date	Feb 20, 2024

Abstract

Techniques are provided for improving computers as tools for assisting in negotiations. Specifically, techniques are provided for using a trained machine learning system to predict the likelihood that a party to a negotiation intends to comply with terms that are under consideration. In some negotiations, each party of a negotiation may use the techniques described herein to determine terms to offer the other party. In such situations, both parties may be both terms-receiving parties and terms-offering parties. By using a trained machine learning system to predict the intent of a party, the possibility of human bias significantly reduced, allowing proposed terms to be based more on objective facts and predictive indicators rather than the prejudices of the agents that have been delegated the responsibility of proposing terms.

First claim

What is claimed is:

1. A method comprising:
training a machine learning system to predict intent of a target party in a current negotiation;
wherein the training involves feeding the machine learning system:
negotiation data for each of a plurality of prior negotiations between parties that did not include the target party; and
outcome data for the plurality of prior negotiations, wherein the outcome data includes data indicating whether agreed-upon terms from each of the plurality of prior negotiations were complied with;
wherein the outcome data includes:
data for one or more negotiations that have a first label indicating agreed-upon terms were complied with; and
data for one or more negotiations that have a second label indicating agreed-upon terms were not complied with;

wherein, for each of the plurality of prior negotiations, the negotiation data is associated with a corresponding label from a set consisting of the first label and the second label;
after training the machine learning system, feeding current negotiation data to the trained machine learning system to cause the trained machine learning system to generate a first intent score that predicts a likelihood that the target party intends to comply with terms under consideration in the current negotiation;
wherein the current negotiation data includes data about interactions with the target party during the current negotiation; and
determining terms to offer to the target party in the current negotiation based, at least in part, on the first intent score.

2. The method of claim 1 wherein the current negotiation data includes first candidate terms, and the first intent score is for the first candidate terms.

3. The method of claim 2 wherein determining terms to offer includes:
determining whether the first intent score satisfies certain criteria;
responsive to determining that the first intent score does not satisfy the certain criteria, formulating second candidate terms;
causing the trained machine learning system to generate a second intent score by feeding current negotiation data with the second candidate terms to the trained machine learning system;

determining whether the second intent score satisfies the certain criteria; and
responsive to determining that the second intent score satisfies the certain criteria, offering the target party the second candidate terms.

4. The method of claim 3 wherein the certain criteria includes that a corresponding confidence score produced by the trained machine learning system exceeds a particular threshold.

5. The method of claim 1 wherein the negotiation data from prior negotiations, and the current negotiation data, include at least one of:
a party's tone of voice,
a party's choice of words,
the frequency that a party uses certain words,
a party's voice modulation,
a party's time of picking or making a call,
where a party is calling from,
who a party is calling with,
length of pauses before a party answers questions,
whether a party circumvents a question,
types of words used by a party, or
amount of time or rings until a party answers a call.

6. The method of claim 1 wherein the current negotiation data fed to the machine learning system includes information about video captured during the current negotiation.

7. The method of claim 6 wherein the information about video captured includes at least one of:
the target party's facial expressions,
how many times the target party nods their head,
the target party's attentiveness,
where the target party focusses their eyes.

8. The method of claim 1 wherein the current negotiation data fed to the machine learning system includes information derived from two or more of:
a text-based chat conversation with the target party;
one or more email messages from the target party;
a texting conversation with the target party;
a video chat with the target party;
a phone conversation with the target party; and
video captured during a live conversation with the target party.

9. A method comprising:
training a machine learning system to predict intent of a target party in a current negotiation;
wherein the training involves feeding the machine learning system:
negotiation data for each of a plurality of prior interactions with the target party; and
outcome data for the plurality of prior interactions, the outcome data including data that indicates whether the target party complied with commitments made during each of the plurality of prior interactions;
wherein the outcome data includes:
data for one or more negotiations that have a first label indicating agreed-upon terms were complied with; and
data for one or more negotiations that have a second label indicating agreed-upon terms were not complied with;

wherein, for each of the plurality of prior negotiations, the negotiation data is associated with a corresponding label from a set consisting of the first label and the second label;
after training the machine learning system, feeding current negotiation data to the trained machine learning system to cause the trained machine learning system to generate a first intent score that predicts a likelihood that the target party will comply with commitments of the target party during the current negotiation;
wherein the current negotiation data includes data about interactions with the target party during the current negotiation; and
determining terms to offer to the target party in the current negotiation based, at least in part, on the first intent score.

10. The method of claim 9 wherein the negotiation data from prior negotiations, and the current negotiation data, include at least one of:
a party's tone of voice,
a party's choice of words,
the frequency that a party uses certain words,
a party's voice modulation,
a party's time of picking or making a call,
where a party is calling from,
who a party is calling with,
length of pauses before a party answers questions,
whether a party circumvents a question,
types of words used by a party, or
amount of time or rings until a party answers a call.

11. The method of claim 9 wherein the current negotiation data fed to the machine learning system includes information about video captured during the current negotiation.

12. The method of claim 11 wherein the information about video captured includes at least one of:
the target party's facial expressions,
how many times the target party nods their head,
the target party's attentiveness,
where the target party focusses their eyes.

13. The one or more non-transitory computer-readable media of claim 9 wherein the current negotiation data includes first candidate terms, and the first intent score is for the first candidate terms.

14. The one or more non-transitory computer-readable media of claim 13 wherein determining terms to offer includes:
determining whether the first intent score satisfies certain criteria;
responsive to determining that the first intent score does not satisfy the certain criteria, formulating second candidate terms;
causing the trained machine learning system to generate a second intent score by feeding current negotiation data with the second candidate terms to the trained machine learning system;

determining whether the second intent score satisfies the certain criteria; and
responsive to determining that the second intent score satisfies the certain criteria, offering the target party the second candidate terms.

15. The one or more non-transitory computer-readable media of claim 9 wherein the current negotiation data fed to the machine lea

Assignees

• Lendingclub Corp

Inventors

• Alagianambi Ramaswamy

Classifications

• G06N20/20 Primary

  Ensemble learning · CPC title

• G06V10/776

  Validation; Performance evaluation · CPC title

• G10L17/26

  Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices · CPC title

• G10L25/63

  for estimating an emotional state · CPC title

• G06F40/30

  Semantic analysis · CPC title