# US 2018/0025281 A1 · Cognitive computing for servers and mobile devices > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2018/0025281 A1 - Document ID: US20180025281A1 - Status: Published application - Kind: A1 - Application number: 15/223,268 - Filed: 2016-07-29 - Published: 2018-01-25 - Earliest claimed: 2016-07-19 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Pradip Bose; Alper Buyuktosunoglu; Augusto Javier Vega - Web page: https://patentgrep.com/patent/US20180025281A1 ## CPC - G06N 20/00: Machine learning - G06N 5/04: Inference or reasoning models - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Techniques for implementing a safety protocol are provided. In one example, a system is provided that can execute a machine-learned model to determine cognitive data representing a prediction about a state of an environment and an action to be performed in response to the prediction. The system can determine that a connection with a remote device is unavailable, and in response activate a safety protocol. ## Claims ### Claim 1 A system, comprising: a memory that stores computer executable components; a processor that executes computer executable components stored in the memory, wherein the computer executable components comprise: a cognitive agent component that, based on context data that represents a state of an environment and based on model data indicative of a machine-learned model of interpreting the context data, determines cognitive data comprising prediction data and action data, wherein the prediction data is indicative of a prediction relating to the state of the environment and the action data is indicative of a first action to be performed in response to the prediction; a communication component that determines failed connection data that indicates communication with a remote device is unavailable; and a safety component that instructs the cognitive agent component to utilize a safety protocol in response to a determination of the failed connection data. ### Claim 2 (depends on 1) The system of claim 1, wherein the cognitive agent component further updates the action data with a second action to be performed in response to the prediction, wherein the second action is determined based on the safety protocol. ### Claim 3 (depends on 1) The system of claim 1, wherein the cognitive agent component further determines confidence data indicative of a confidence measure of the prediction, wherein the cognitive data comprises the confidence data. ### Claim 4 (depends on 3) The system of claim 3, wherein the safety component instructs the cognitive agent component to utilize the safety protocol to determine the cognitive data in response to the determination of the failed connection data and further in response to a second determination that the confidence measure fails to meet a defined confidence threshold. ### Claim 5 (depends on 1) The system of claim 1, wherein the cognitive agent component further determines severity data indicative of a severity metric associated with the action, wherein the cognitive data comprises the severity data. ### Claim 6 (depends on 5) The system of claim 5, wherein the safety component instructs the cognitive agent component to utilize the safety protocol to determine the cognitive data in response to the determination of the failed connection data and further in response to a second determination that the determined severity fails to meet a defined severity threshold. ### Claim 7 (depends on 1) The system of claim 1, wherein the cognitive agent component updates the action data according to the safety protocol and determines a second action to be performed in response to the prediction, and wherein the second action is determined to provide additional context data employed to improve a confidence measure of the prediction. ### Claim 8 (depends on 1) The system of claim 1, wherein the cognitive agent component updates the action data according to the safety protocol and determines a second action to be performed in response to the prediction, wherein the first action is determined to have a first severity metric and the second action is determined to have a second severity metric, and wherein the first severity metric is greater than the second severity metric. ### Claim 9 (depends on 1) The system of claim 1, wherein the communication component further determines connection data that indicates communication with the remote device is available. ### Claim 10 (depends on 9) The system of claim 9, wherein the safety component, in response to determination of the connection data, instructs the cognitive agent component to generate the cognitive data without employing the safety protocol. ### Claim 11 (depends on 1) The system of claim 1, further comprising a sensor component that generates the context data in response to input received from the environment. ### Claim 12 (depends on 1) The system of claim 1, wherein the system is included in an unmanned vehicle. ### Claim 13 A computer program product for interpreting context information, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to: determine, by the processor, cognitive data comprising prediction data that represents a prediction relating to the state of the environment and action data that represents a first action to be performed in response to the prediction, wherein determination of the cognitive data is based on context data that represents a state of an environment and based on model data that describes a machine-learned model of interpreting the context data, wherein the model comprises a safety protocol; determine, by the processor, connection unavailability data that represents an indication that communication with a remote device that trains the model is unavailable; and determine, by the processor, the cognitive data based on the safety protocol of the model data, wherein determination of the cognitive data based on the safety protocol is performed in response to determination of the connection unavailability data. ### Claim 14 (depends on 13) The computer program product of claim 13, wherein the program instructions are further executable to determine, by the processor, an update to the action data comprising a second action to be performed in response to the prediction, wherein the second action is determined based on the safety protocol. ### Claim 15 (depends on 14) The computer program product of claim 14, wherein the second action is determined to provide additional context data employed to improve a confidence measure of the prediction, wherein improvement of the confidence measure is based on analysis of the additional context data. ### Claim 16 (depends on 14) The computer program product of claim 14, wherein the first action is associated with a first severity metric representing a determined severity of the first action and the second action is associated with a second severity metric representing a determined severity of the second action, and wherein the second severity metric is less than the first severity metric. ### Claim 17 A computer-implemented method, comprising: determining, by a device operatively coupled to a processor, cognitive data comprising prediction data, confidence data, and action data, wherein the prediction data represents a prediction relating to a characteristic of an environment, wherein the confidence data represents a confidence measure of the prediction, wherein the action data represents an action to be performed in response to the prediction, and wherein the determining is based on context data and model data, the context data representing the characteristic of the environment and model data represents a machine-learned model of interpreting the context data; determining, by the device, connection unavailability data representing an indication that communication with a remote device that trains a safety protocol of the model data is unavailable; and in response to the determining the connection unavailability data, generating, by the device, the cognitive data based on the safety protocol. ### Claim 18 (depends on 17) The computer-implemented method of claim 17, further comprising determining, by the device, an update to the action data comprising a second action to be performed in response to the prediction, wherein the second action is determined based on the safety protocol. ### Claim 19 (depends on 17) The computer-implemented method of claim 17, further comprising determining, by the device, connection availability data representing an indication that communication with the remote device is available. ### Claim 20 (depends on 19) The computer-implemented method of claim 19, further comprising, in response to the determining the connection availability data, determining, by the device, the cognitive data based on the model data exclusive of the safety protocol.