# US 2018/0322402 A1 · Method and system for a user-specific cognitive unit that enhances generic recommendation systems

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2018/0322402 A1
- Document ID: US20180322402A1
- Status: Published application
- Kind: A1
- Application number: 15/970,877
- Filed: 2018-05-04
- Published: 2018-11-08
- Earliest claimed: 2017-05-03
- Assignees: Gluru Limited
- Applicants: Gluru Limited
- Inventors: Ritwik Kulkarni; Ozgun Tandiroglu; Michele Sama; Tim Porter
- Web page: https://patentgrep.com/patent/US20180322402A1

## CPC

- G06N 5/04: Inference or reasoning models
- G06F 9/451: Execution arrangements for user interfaces
- G06F 9/461: {Saving or restoring of program or task context}
- G06F 11/3438: {monitoring of user actions (tracking the activity of the user H04L67/535)}
- G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method and system are disclosed for a novel architecture in which competing suggestions, possibly generated by competing systems, are selected by a Cognitive Unit (CU). The CU observes the user context and learns which contextual circumstances affect the user's cognitive behaviour. Majority of the traditional models trained over multiple users fail to represent the individual because (1) they ignore personal bias toward certain decisions and (2) they don't have complete visibility of all options available to users (i.e. from competitive systems). The invention is ideally suited to interact with several other products as more and more modern products are using AIs to drive the user experience. That shifts traditional HCI towards a novel form of interaction that we call human-AI interaction (HAII). When applied to user experience, predictive models make decisions on users' behalf attempting to minimise user interaction while guiding them toward the completion of predefined funnels. This invention presents a novel approach using a personal model capable of partially replicating cognitive behaviours by learning personal situational biases using the CU. By leveraging the personal biases, the system allows the AI to learn situational biases otherwise not available to predictive models. Once trained cognitive unit can supplement the decision-making process of the user.

## Claims

### Claim 1

A computer-implemented method for human artificial intelligence interaction (HAII) comprising:
displaying one or more suggested tasks to a user;
receiving a first input from the user, the first input indicating a selected suggested task from the one or more suggested tasks;
in response to receiving the first input from the user:
training a cognitive unit with the first input from the user so that the cognitive unit learns a bias of the user associated with the suggested task;
displaying to the user a predicted action associated with the selected suggested task; and
receiving a second input from the user, the second input indicating whether the predicted action is accepted by the user;
in response to the second input from the user indicating that the predicted action is accepted by the user:
training a cognitive unit with the second input from the user so that the cognitive unit learns the predicted action is acceptable to the user for the selected suggested task; and
saving the suggested task and the predicted action in a task manager;
in response to the second input from the user indicating that the predicted action is not accepted by the user:
training a cognitive unit with the second input from the user so that the cognitive unit learns the predicted action is not acceptable to the user for the selected suggested task;
displaying one or more situationally suitable actions to the user, the situationally suitable actions prioritized by the cognitive unit; and
receiving a third input from the user, the third input second input indicating a selected situationally suitable action from the one or more situationally suitable actions;
training a cognitive unit with the third input from the user so that the cognitive unit learns the selected situationally suitable action is acceptable to the user for the selected suggested task.

### Claim 2 (depends on 1)

The method of claim 1, in which displaying one or more suggested tasks to a user includes:
receiving a trigger input,
displaying the one or more suggested tasks to a user based on the trigger input; and
training the cognitive unit with the trigger input.

### Claim 3 (depends on 1)

The method of claim 1, in which displaying one or more suggested tasks to a user includes:
receiving a context,
displaying the one or more suggested tasks to a user based on the context; and
training the cognitive unit with the context.

### Claim 4 (depends on 1)

The method of claim 1, in which the predicted action is received from a global predictive model of a third-party system.

### Claim 5 (depends on 4)

The method of claim 4, in which a plurality of predicted actions are received from a plurality of global predictive models of a plurality of third-party systems, and the cognitive unit selects the predicted action to display to the user.

### Claim 6 (depends on 1)

The method of claim 1, in which the cognitive unit learns over time using reinforcement learning, on-line learning, or mini batches.

### Claim 7

A computer system for human artificial intelligence interaction (HAII) comprising:
a computer processor;
a computer-readable memory coupled to the computer processor, the computer readable memory encoded with computer instructions that, when executed by the computer processor, cause the computer system to performs the steps of:
displaying one or more suggested tasks to a user;
receiving a first input from the user, the first input indicating a selected suggested task from the one or more suggested tasks;
in response to receiving the first input from the user:
training a cognitive unit with the first input from the user so that the cognitive unit learns a bias of the user associated with the suggested task;
displaying to the user a predicted action associated with the selected suggested task; and
receiving a second input from the user, the second input indicating whether the predicted action is accepted by the user;
in response to the second input from the user indicating that the predicted action is accepted by the user:
training a cognitive unit with the second input from the user so that the cognitive unit learns the predicted action is acceptable to the user for the selected suggested task; and
saving the suggested task and the predicted action in a task manager;
in response to the second input from the user indicating that the predicted action is not accepted by the user:
training a cognitive unit with the second input from the user so that the cognitive unit learns the predicted action is not acceptable to the user for the selected suggested task;
displaying one or more situationally suitable actions to the user, the situationally suitable actions prioritized by the cognitive unit; and
receiving a third input from the user, the third input second input indicating a selected situationally suitable action from the one or more situationally suitable actions;
training a cognitive unit with the third input from the user so that the cognitive unit learns the selected situationally suitable action is acceptable to the user for the selected suggested task.

### Claim 8 (depends on 7)

The system of claim 7, in which displaying one or more suggested tasks to a user includes:
receiving a trigger input,
displaying the one or more suggested tasks to a user based on the trigger input; and
training the cognitive unit with the trigger input.

### Claim 9 (depends on 7)

The system of claim 7, in which displaying one or more suggested tasks to a user includes:
receiving a context,
displaying the one or more suggested tasks to a user based on the context; and
training the cognitive unit with the context.

### Claim 10 (depends on 7)

The system of claim 7, in which the predicted action is received from a global predictive model of a third-party system.

### Claim 11 (depends on 10)

The system of claim 10, in which a plurality of predicted actions are received from a plurality of global predictive models of a plurality of third-party systems, and the cognitive unit selects the predicted action to display to the user.

### Claim 12 (depends on 1)

The system of claim 1, in which the cognitive unit learns over time using reinforcement learning, on-line learning, or mini batches.

### Claim 13

A non-transitory computer-readable medium encoded with computer instructions for human artificial intelligence interaction (HAII), the computer instructions, when executed by a computer system, cause the computer system to perform the steps of:
displaying one or more suggested tasks to a user;
receiving a first input from the user, the first input indicating a selected suggested task from the one or more suggested tasks;
in response to receiving the first input from the user:
training a cognitive unit with the first input from the user so that the cognitive unit learns a bias of the user associated with the suggested task;
displaying to the user a predicted action associated with the selected suggested task; and
receiving a second input from the user, the second input indicating whether the predicted action is accepted by the user;
in response to the second input from the user indicating that the predicted action is accepted by the user:
training a cognitive unit with the second input from the user so that the cognitive unit learns the predicted action is acceptable to the user for the selected suggested task; and
saving the suggested task and the predicted action in a task manager;
in response to the second input from the user indicating that the predicted action is not accepted by the user:
training a cognitive unit with the second input from the user so that the cognitive unit learns the predicted action is not acceptable to the user for the selected suggested task;
displaying one or more situationally suitable actions to the user, the situationally suitable actions prioritized by the cognitive unit; and
receiving a third input from the user, the third input second input indicating a selected situationally suitable action from the one or more situationally suitable actions;
training a cognitive unit with the third input from the user so that the cognitive unit learns the selected situationally suitable action is acceptable to the user for the selected suggested task.

### Claim 14 (depends on 13)

The non-transitory computer-readable medium of claim 13, in which displaying one or more suggested tasks to a user includes:
receiving a trigger input,
displaying the one or more suggested tasks to a user based on the trigger input; and
training the cognitive unit with the trigger input.

### Claim 15 (depends on 13)

The non-transitory computer-readable medium of claim 13, in which displaying one or more suggested tasks to a user includes:
receiving a context,
displaying the one or more suggested tasks to a user based on the context; and
training the cognitive unit with the context.

### Claim 16 (depends on 13)

The non-transitory computer-readable medium of claim 13, in which the predicted action is received from a global predictive model of a third-party system.

### Claim 17 (depends on 16)

The non-transitory computer-readable medium of claim 16, in which a plurality of predicted actions are received from a plurality of global predictive models of a plurality of third-party systems, and the cognitive unit selects the predicted action to display to the user.

### Claim 18 (depends on 13)

The non-transitory computer-readable medium of claim 13, in which the cognitive unit learns over time using reinforcement learning, on-line learning, or mini batches.