Artificial intelligence (AI) technologies offer tremendous potential for product and service innovation, yet finding good use cases remains challenging. Currently, AI projects largely fail due to breakdowns in early stage ideation and problem formulation. Drawing on HCI research that used AI capabilities and examples to facilitate AI concept ideation, this paper investigates how these approaches might be operationalized in industry settings. We collaborated with cross-functional industry teams in insurance, accounting, and consultancy. We conducted a series of AI Design Sprints, where innovators simultaneously consider AI capabilities and human needs. All teams perceived the ideation method highly valuable both for rapidly exploring use cases and building AI literacy within teams. We detail our process, the challenges, and artifacts that proved effective. We share insights on how AI projects get initiated, and how innovation teams identify use cases. Reflecting on these case studies, we discuss opportunities for improving early stage AI innovation.

Whose labels should a machine learning (ML) algorithm learn to emulate? For ML tasks ranging from online comment toxicity to misinformation detection to medical diagnosis, different groups in society may have irreconcilable disagreements about ground truth labels. Supervised ML today resolves these label disagreements implicitly using majority vote, which overrides minority groups’ labels. We introduce jury learning, a supervised ML approach that resolves these disagreements explicitly through the metaphor of a jury: defining which people or groups, in what proportion, determine the classifier’s prediction. For example, a jury learning model for online toxicity might centrally feature women and Black jurors, who are commonly targets of online harassment. To enable jury learning, we contribute a deep learning architecture that models every annotator in a dataset, samples from annotators’ models to populate the jury, then runs inference to classify. Our architecture enables juries that dynamically adapt their composition, explore counterfactuals, and visualize dissent. A field evaluation finds that practitioners construct diverse juries that alter 14% of classification outcomes.

The confusion matrix, a ubiquitous visualization for helping people evaluate machine learning models, is a tabular layout that compares predicted class labels against actual class labels over all data instances. We conduct formative research with machine learning practitioners at Apple and find that conventional confusion matrices do not support more complex data-structures found in modern-day applications, such as hierarchical and multi-output labels. To express such variations of confusion matrices, we design an algebra that models confusion matrices as probability distributions. Based on this algebra, we develop Neo, a visual analytics system that enables practitioners to flexibly author and interact with hierarchical and multi-output confusion matrices, visualize derived metrics, renormalize confusions, and share matrix specifications. Finally, we demonstrate Neo’s utility with three model evaluation scenarios that help people better understand model performance and reveal hidden confusions.

Machine learning classifiers for human-facing tasks such as comment toxicity and misinformation often score highly on metrics such as ROC AUC but are received poorly in practice. Why this gap? Today, metrics such as ROC AUC, precision, and recall are used to measure technical performance; however, human-computer interaction observes that evaluation of human-facing systems should account for people’s reactions to the system. In this paper, we introduce a transformation that more closely aligns machine learning classification metrics with the values and methods of user-facing performance measures. The disagreement deconvolution takes in any multi-annotator (e.g., crowdsourced) dataset, disentangles stable opinions from noise by estimating intra-annotator consistency, and compares each test set prediction to the individual stable opinions from each annotator. Applying the disagreement deconvolution to existing social computing datasets, we find that current metrics dramatically overstate the performance of many human-facing machine learning tasks: for example, performance on a comment toxicity task is corrected from .95 to .73 ROC AUC.

Successful machine learning (ML) applications require iterations on both modeling and the underlying data. While prior visualization tools for ML primarily focus on modeling, our interviews with 23 ML practitioners reveal that they improve model performance frequently by iterating on their data (e.g., collecting new data, adding labels) rather than their models. We also identify common types of data iterations and associated analysis tasks and challenges. To help attribute data iterations to model performance, we design a collection of interactive visualizations and integrate them into a prototype, Chameleon, that lets users compare data features, training/testing splits, and performance across data versions. We present two case studies where developers apply \system to their own evolving datasets on production ML projects. Our interface helps them verify data collection efforts, find failure cases stretching across data versions, capture data processing changes that impacted performance, and identify opportunities for future data iterations.

Analyzing queries from search engines and intelligent assistants is difficult. A key challenge is organizing queries into interpretable, context-preserving, representative, and flexible groups. We present structural templates, abstract queries that replace tokens with their linguistic feature forms, as a query grouping method. The templates allow analysts to create query groups with structural similarity at different granularities. We introduce Tempura, an interactive tool that lets analysts explore a query dataset with structural templates. Tempura summarizes a query dataset by selecting a representative subset of templates to show the query distribution. The tool also helps analysts navigate the template space by suggesting related templates likely to yield further explorations. Our user study shows that Tempura helps analysts examine the distribution of a query dataset, find labeling errors, and discover model error patterns and outliers.

We aim to increase the flexibility at which a data worker can choose the right tool for the job, regardless of whether the tool is a code library or an interactive graphical user interface (GUI). To achieve this flexibility, we extend computational notebooks with a new API mage, which supports tools that can represent themselves as both code and GUI as needed. We discuss the design of mage as well as design opportunities in the space of flexible code/GUI tools for data work. To understand tooling needs, we conduct a study with nine professional practitioners and elicit their feedback on mage and potential areas for flexible code/GUI tooling. We then implement six client tools for mage that illustrate the main themes of our study findings. Finally, we discuss open challenges in providing flexible code/GUI interactions for data workers.

We have limited understanding of how older adults use smartphones, how their usage differs from younger users, and the causes for those differences. As a result, researchers and developers may miss promising opportunities to support older adults or offer solutions to unimportant problems. To characterize smartphone usage among older adults, we collected iPhone usage data from 84 healthy older adults over three months. We find that older adults use fewer apps, take longer to complete tasks, and send fewer messages. We use cognitive test results from these same older adults to then show that up to 79% of these differences can be explained by cognitive decline, and that we can predict cognitive test performance from smartphone usage with 83% ROCAUC. While older adults differ from younger adults in app usage behavior, the "cognitively young" older adults use smartphones much like their younger counterparts. Our study suggests that to better support all older adults, researchers and developers should consider the full spectrum of cognitive function.

The majority of machine learning research has been focused on building models and inference techniques with sound mathematical properties and cutting edge performance. Little attention has been devoted to the development of data representation that can be used to improve a user's ability to interpret the data and machine learning models to solve real-world problems. In this paper, we quantitatively and qualitatively evaluate an efficient, accurate and scalable feature-compression method using latent Dirichlet allocation for discrete data. This representation can effectively communicate the characteristics of high-dimensional, complex data points. We show that the improvement of a user's interpretability through the use of a topic modeling-based compression technique is statistically significant, according to a number of metrics, when compared with other representations. Also, we find that this representation is scalable --- it maintains alignment with human classification accuracy as an increasing number of data points are shown. In addition, the learned topic layer can semantically deliver meaningful information to users that could potentially aid human reasoning about data characteristics in connection with compressed topic space.

Google Research recently tested a massive online class model for an internal engineering education program, with machine learning as the topic, that blended theoretical concepts and Google-specific software tool tutorials. The goal of this training was to foster engineering capacity to leverage machine learning tools in future products. The course was delivered both synchronously and asynchronously, and students had the choice between studying independently or participating with a group. Since all students are company employees, unlike most publicly offered MOOCs we can continue to measure the students' behavioral change long after the course is complete. This paper describes the course, outlines the available data set and presents directions for analysis.

We present Gestalt, a development environment designed to support the process of applying machine learning. While traditional programming environments focus on source code, we explicitly support both code and data. Gestalt allows developers to implement a classification pipeline, analyze data as it moves through that pipeline, and easily transition between implementation and analysis. An experiment shows this significantly improves the ability of developers to find and fix bugs in machine learning systems. Our discussion of Gestalt and our experimental observations provide new insight into general-purpose support for the machine learning process.

Although existing work has explored both information extraction and community content creation, most research has focused on them in isolation. In contrast, we see the greatest leverage in the synergistic pairing of these methods as two interlocking feedback cycles. This paper explores the potential synergy promised if these cycles can be made to accelerate each other by exploiting the same edits to advance both community content creation and learning-based information extraction. We examine our proposed synergy in the context of Wikipedia infoboxes and the Kylin information extraction system. After developing and refining a set of interfaces to present the verification of Kylin extractions as a non primary task in the context of Wikipedia articles, we develop an innovative use of Web search advertising services to study people engaged in some other primary task. We demonstrate our proposed synergy by analyzing our deployment from two complementary perspectives: (1) we show we accelerate community content creation by using Kylin's information extraction to significantly increase the likelihood that a person visiting a Wikipedia article as a part of some other primary task will spontaneously choose to help improve the article's infobox, and (2) we show we accelerate information extraction by using contributions collected from people interacting with our designs to significantly improve Kylin's extraction performance.

As statistical machine learning algorithms and techniques continue to mature, many researchers and developers see statistical machine learning not only as a topic of expert study, but also as a tool for software development. Extensive prior work has studied software development, but little prior work has studied software developers applying statistical machine learning. This paper presents interviews of eleven researchers experienced in applying statistical machine learning algorithms and techniques to human-computer interaction problems, as well as a study of ten participants working during a five-hour study to apply statistical machine learning algorithms and techniques to a realistic problem. We distill three related categories of difficulties that arise in applying statistical machine learning as a tool for software development: (1) difficulty pursuing statistical machine learning as an iterative and exploratory process, (2) difficulty understanding relationships between data and the behavior of statistical machine learning algorithms, and (3) difficulty evaluating the performance of statistical machine learning algorithms and techniques in the context of applications. This paper provides important new insight into these difficulties and the need for development tools that better support the application of statistical machine learning.

Many mobile machine learning applications require collecting and labeling data, and a traditional GUI on a mobile device may not be an appropriate or viable method for this task. This paper presents an alternative approach to mobile labeling of sensor data called VoiceLabel. VoiceLabel consists of two components: (1) a speech-based data collection tool for mobile devices, and (2) a desktop tool for offline segmentation of recorded data and recognition of spoken labels. The desktop tool automatically analyzes the audio stream to find and recognize spoken labels, and then presents a multimodal interface for reviewing and correcting data labels using a combination of the audio stream, the system's analysis of that audio, and the corresponding mobile sensor data. A study with ten participants showed that VoiceLabel is a viable method for labeling mobile sensor data. VoiceLabel also illustrates several key features that inform the design of other data labeling tools.

Navigation services (e.g., in-car navigation systems and online mapping sites) compute routes between two locations to help users navigate. However, these routes may direct users along an unfamiliar path when a familiar path exists, or, conversely, may include redundant information that the user already knows. These overly complicated directions increase the cognitive load of the user, which may lead to a dangerous driving environment. Since the level of detail is user specific and depends on their familiarity with a region, routes need to be personalized. We have developed a system, called MyRoute, that reduces route complexity by creating user specific routes based on a priori knowledge of familiar routes and landmarks. MyRoute works by compressing well known steps into a single contextualized step and rerouting users along familiar routes.

In this paper we propose an active control strategy for scanning laser sensors on autonomous vehicles traveling offroad at high speeds. As speed increases the amount of sensor information about the terrain decreases. We address the problem of sensor control in the context of this speed-coverage trade off. The algorithm and testing methodologies are described with results comparing our active sensing method to a passive sensing method.