‍

Aggregation:

• Definition: Combining data from multiple sources or time periods to reduce the risk of individual identification.
• Kiwibot Use Case: Combining data from multiple Kiwibot robots in different campuses and environmental conditions to analyze trends in delivery patterns, identify areas for improvement, and optimize operations.

Augmentation:

• Definition: Enhancing human capabilities or potential through technology while preserving human agency.
• Kiwibot Use Case: Enhancing human capabilities by automating repetitive tasks, such as order processing and route planning, allowing human operators to focus on more strategic activities.

‍

Automation:

• Definition: Performing tasks without human involvement using machines or software.
• Kiwibot Use Case: Automating delivery processes, including order fulfillment, route planning, and navigation, to improve efficiency and reduce costs.

Artificial Intelligence (AI):

• Definition: AI refers to the development of computer systems that can perform tasks typically requiring human intelligence.
• Kiwibot Use Case: Kiwibot's robots leverage AI algorithms for tasks like navigation, object recognition, and decision-making.

Autonomous Navigation:

• Definition: The ability of a robot to navigate independently without human intervention.
• Kiwibot Use Case: Kiwibot's robots are designed for autonomous navigation, using AI and robotics technologies to complete delivery tasks.

Binary Classification:

• Definition: An ML model that predicts whether an example belongs to one of two categories.
• Kiwibot Use Case: Classifying customer feedback as positive or negative to identify areas for improvement and optimize the user experience.

Classification:

• Definition: Identifying an object using ML.
• Kiwibot Use Case: Classifying objects in the environment, such as pedestrians, vehicles, and obstacles, to enable safe and efficient navigation.

Computer Vision:

• Definition: A field of AI that deals with enabling computers to understand visual information.
• Kiwibot Use Case: Kiwibot's robots use computer vision to process images, identify objects, and navigate their surroundings.

Confidence Level:

• Definition: A statistical measure of how certain a prediction or outcome is.
• Kiwibot Use Case: Measuring the confidence level of Kiwibot's models in their predictions, such as the likelihood of a successful delivery or the accuracy of obstacle detection.

Context Errors:

• Definition: Situations where system output is not relevant to the user's current situation.
• Kiwibot Use Case: Identifying and addressing situations where Kiwibot's actions are not relevant to the current context, such as delivering an order to the wrong location or attempting to navigate an inaccessible area.

Counterfactuals:

• Definition: Explanations for why something is not classified as belonging to a given class, often presented as hypothetical scenarios.
• Kiwibot Use Case: Using counterfactual reasoning to evaluate the impact of different decisions or actions on delivery outcomes, such as analyzing what would have happened if a robot had taken a different route or encountered different obstacles.

Data Cascades:

• Definition: Negative downstream effects caused by data issues, leading to technical debt.
• Kiwibot Use Case: Identifying and addressing data-related issues that can have cascading effects on Kiwibot's operations, such as errors in location data or incorrect order information.

Data Collection and Labeling:

• Definition: Acquiring and labeling data for ML models.
• Kiwibot Use Case: Collecting and labeling data from Kiwibot's sensors and interactions with the environment to train and improve machine learning models for tasks like object detection, path planning, and customer satisfaction analysis.

Data Distribution:

• Definition: The frequency of specific values within a dataset.
• Kiwibot Use Case: Analyzing the distribution of data related to delivery orders, traffic patterns, and customer behavior to identify trends and optimize operations.

Data Examples:

• Definition: Individual pieces of data, such as images or text.
• Kiwibot Use Case: Images captured by Kiwibot's cameras, GPS data, sensor readings, customer feedback, and order information.

Data Features:

• Definition: Measurable properties of data, used as input for ML models.
• Kiwibot Use Case: Extracting relevant features from data, such as object locations, distances, and environmental conditions, to train machine learning models for tasks like obstacle avoidance and route optimization.

Data Labels:

• Definition: Human-assigned descriptions for data examples.
• Kiwibot Use Case: Assigning labels to data examples, such as classifying objects in images as pedestrians, vehicles, or obstacles, or labeling customer feedback as positive or negative.

Deep Learning:

• Definition: A type of ML using artificial neural networks with multiple layers.
• Kiwibot Use Case: Kiwibot employs deep learning for tasks like image recognition, natural language processing, and sensor fusion.

Explicit Data Collection:

• Definition: Directly requesting information from users.
• Kiwibot Use Case: Collecting explicit feedback from customers through surveys or in-app ratings to gather information about their experiences and identify areas for improvement.

Explicit Feedback:

• Definition: User-provided information through in-app mechanisms like ratings or surveys.
• Kiwibot Use Case: Analyzing customer feedback to identify areas for improvement, such as delivery speed, accuracy, and customer service.

False Negatives:

• Definition: Incorrectly classifying an object as not belonging to a category.
• Kiwibot Use Case: Identifying instances where Kiwibot's models incorrectly classify objects or situations, such as failing to detect an obstacle or misinterpreting a customer's request.

False Positives:

• Definition: Incorrectly classifying an object as belonging to a category.
• Kiwibot Use Case: Identifying instances where Kiwibot's models incorrectly classify objects or situations, such as mistaking a shadow for a pedestrian or delivering an order to the wrong location.

Features:

• Definition: Distinct data sources or calculations that influence predictions.
• Kiwibot Use Case: Utilizing various features extracted from data, such as object locations, distances, and environmental conditions, to train machine learning models for tasks like obstacle avoidance and route optimization.

Folk Theories:

• Definition: Inaccurate user beliefs about how a product works.
• Kiwibot Use Case: Identifying and addressing misconceptions that users may have about Kiwibot's capabilities and limitations.

General System Explanations:

• Definition: Descriptions of system functionality and how inputs lead to outputs.
• Kiwibot Use Case: Providing clear and concise explanations of Kiwibot's system architecture, components, and how they work together to deliver autonomous delivery services.

Heuristic-Based:

• Definition: Rule-based systems that produce specific results based on predefined conditions.
• Kiwibot Use Case: Using heuristic rules to guide decision-making in certain situations, such as avoiding specific areas or following predefined routes.

Implicit Data Collection:

• Definition: Gathering information about users passively through behavior logging.
• Kiwibot Use Case: Collecting implicit feedback from user interactions, such as delivery times, order cancellations, and customer ratings.

Implicit Feedback:

• Definition: Information inferred from user interactions within an application.
• Kiwibot Use Case: Analyzing implicit feedback to identify user preferences, satisfaction levels, and areas for improvement.

Inter-labeler Reliability:

• Definition: Consistency among different labelers performing the same task.
• Kiwibot Use Case: Ensuring consistency and reliability in the labeling process by evaluating agreement among different human labelers.

Labeler:

• Definition: A person who labels data for ML training.
• Kiwibot Use Case: Human experts who label data for Kiwibot's machine learning models.

Labeled Data:

• Definition: Data that has been annotated with human-generated labels.
• Kiwibot Use Case: Data that has been annotated with human-generated labels, such as object classifications or route information.

Localization and Mapping (SLAM):

• Definition: Simultaneously building a map and tracking a robot's position within it.
• Kiwibot Use Case: Kiwibot's robots use SLAM to navigate efficiently and avoid obstacles.

ML Model:

• Definition: A mathematical algorithm that learns from data to make predictions.
• Kiwibot Use Case: Machine learning models are used for various tasks, such as object detection, path planning, and customer satisfaction analysis.

Machine Learning (ML):

• Definition: Training algorithms on large datasets to identify patterns and make predictions. Techniques for programming computers to learn from data and adapt to new situations.
• Kiwibot Use Case: Kiwibot uses ML techniques to train its robot models to learn from data, improve performance, and adapt to new situations.

Machine Learning Systems:

• Definition: Systems that use ML techniques to develop AI.
• Kiwibot Use Case: The underlying systems that power Kiwibot's autonomous delivery capabilities, including the hardware, software, and algorithms that enable the robots to perceive their environment, make decisions, and navigate safely.

Mental Model:

• Definition: Users' internal explanations of how a product works.
• Kiwibot Use Case: Understand how users perceive and interact with Kiwibot's services and design the user experience accordingly.

N-Best Classifications:

• Definition: Showing a specified number of top solutions or suggestions.
• Kiwibot Use Case: Help 

Network Effect:

• Definition: The influence of a product's popularity on user adoption.
• Kiwibot Use Case: Leveraging the network effect to increase user adoption and utilization of Kiwibot's services by partnering with local businesses and promoting the benefits of autonomous delivery.

Obstacle Detection and Avoidance:

• Definition: Identifying potential obstacles and taking action to avoid them.
• Kiwibot Use Case: Kiwibot's robots use a combination of sensors, cameras, and algorithms to detect obstacles and plan safe navigation paths.

Overfitting:

• Definition: A model that performs well on training data but poorly on new data.
• Kiwibot Use Case: Avoid overfitting by ensuring that Kiwibot's models generalize well to new data and do not become too specialized to the training data.

Partial Explanations:

• Definition: Explanations of one aspect of a system's functionality.
• Kiwibot Use Case: Providing users with partial explanations

Path Planning:

• Definition: Determining the optimal route for a robot to follow.
• Kiwibot Use Case: Kiwibot's robots use path-planning algorithms to navigate efficiently and avoid obstacles while minimizing travel time.

Sensor Fusion:

• Definition: Combining data from multiple sensors for a more comprehensive understanding of the environment.
• Kiwibot Use Case: Kiwibot's robots use sensor fusion to integrate data from LiDAR, cameras, and other sensors to improve their perception and navigation capabilities.

‍