Abstract: An implicit bias monitoring system is provided. The system receives sensor data including video and audio that are captured during an interaction between a first individual and a second individual. The system determines a set of attributes associated with the interaction based on the received sensor data, the set of attributes comprising at least one of race and gender of the second individual. The system detects an implicit bias based on a combination of the determined set of attributes and a bias profile. The system generates an alert based on the detected implicit bias.

Abstract: Interpretation maps of deep neural networks are provided that use Renyi differential privacy to guarantee the robustness of the interpretation. In one aspect, a method for generating interpretation maps with guaranteed robustness includes: perturbing an original digital image by adding Gaussian noise to the original digital image to obtain m noisy images; providing the m noisy images as input to a deep neural network; interpreting output from the deep neural network to obtain m noisy interpretations corresponding to the m noisy images; thresholding the m noisy interpretations to obtain a top-k of the m noisy interpretations; and averaging the top-k of the m noisy interpretations to produce an interpretation map with certifiable robustness.

Abstract: An automated analytic tool (AAT) apparatus analyzes a machine learning system (MLS). The tool comprises a processor configured to receive experiment parameters associated with an experiment performed on the MLS, and captures information from a plurality of stages of the experiment. The information comprises information regarding MLS results and choices made during the experiment. The tool automatically revise the captured information into revised information utilizing a knowledge base comprising information from prior experiments. The tool then presents the revised information to a user.

Abstract: Interpretation maps of deep neural networks are provided that use Renyi differential privacy to guarantee the robustness of the interpretation. In one aspect, a method for generating interpretation maps with guaranteed robustness includes: perturbing an original digital image by adding Gaussian noise to the original digital image to obtain m noisy images; providing the m noisy images as input to a deep neural network; interpreting output from the deep neural network to obtain m noisy interpretations corresponding to the m noisy images; thresholding the m noisy interpretations to obtain a top-k of the m noisy interpretations; and averaging the top-k of the m noisy interpretations to produce an interpretation map with certifiable robustness.

Abstract: A method, system, and program product for generating and modifying a video response is provided. The method includes receiving an audio/video file. Parsed video features of the audio/video file are generated with respect to a first graph. Parsed audio features of the audio/video file are generated with respect to a second graph. The first graph is placed overlaying the second graph and at least one intersection point between the first graph and the second graph is determined. A natural language query is executed with respect to the audio/video file and a parsed query entity is generated from the natural language query. The parsed query entity is analyzed with respect to the intersection point and a node of the intersection point comprising similar features is determined with respect to the parsed query entity. A resulting natural language response with respect to the natural language query is generated.

Abstract: An implicit bias monitoring system is provided. The system receives sensor data including video and audio that are captured during an interaction between a first individual and a second individual. The system determines a set of attributes associated with the interaction based on the received sensor data, the set of attributes comprising at least one of race and gender of the second individual. The system detects an implicit bias based on a combination of the determined set of attributes and a bias profile. The system generates an alert based on the detected implicit bias.

Abstract: A method, system, and program product for generating and modifying a video response is provided. The method includes receiving an audio/video file. Parsed video features of the audio/video file are generated with respect to a first graph. Parsed audio features of the audio/video file are generated with respect to a second graph. The first graph is placed overlaying the second graph and at least one intersection point between the first graph and the second graph is determined. A natural language query is executed with respect to the audio/video file and a parsed query entity is generated from the natural language query. The parsed query entity is analyzed with respect to the intersection point and a node of the intersection point comprising similar features is determined with respect to the parsed query entity. A resulting natural language response with respect to the natural language query is generated.