Kevin Crowston’s research while affiliated with Syracuse University and other places

The recent addition of data journalists to several dozen U.S. public radio newsrooms has created multiple new hybridities in the form. No longer are numbers and large datasets “audio poison.” Instead, they are an essential tool for these journalists, who prize journalism’s interpretive function, expressing information in new ways and challenging conventions of broadcast newsroom employment. This study, which relies on semi-structured interviews with 13 public radio data journalists, uses Carlson’s boundary work typology to analyze the ways in which data journalists are expanding the boundaries of U.S. public radio journalism, as well as ways in which they have pushed back against expulsionary pressures. This study’s findings problematize the idea that the results of boundary work must be expressed as in-or-out proposition. Rather, U.S. public radio data journalists suggest their boundaries are a continuum where they may be conditionally accepted by their colleagues, depending on deadlines and on the skills possessed by non-data journalists.

Introduction. Deskilling is a long-standing prediction of the use of information technology, raised anew by the increased capabilities of AI (AI) systems. A review of studies of AI applications suggests that deskilling (or levelling of ability) is a common outcome, but systems can also require new skills, i.e., upskilling. Method. To identify which settings are more likely to yield deskilling vs. upskilling, we propose a model of a human interacting with an AI system for a task. The model highlights the possibility for a worker to develop and exhibit (or not) skills in prompting for, and evaluation and editing of system output, thus yielding upskilling or deskilling. Findings. We illustrate these model-predicted effects on work with examples of current studies of AI-based systems. Conclusions. We discuss organizational implications of systems that deskill or upskill workers and suggest future research directions.

As AI becomes more capable, it is unclear how human creativity will remain essential in jobs that incorporate AI. We conducted a 14-week study of a student newsroom using an AI tool to convert web articles into social media videos. Most treated the tool as a creative springboard, yet still had to edit many AI outputs. The tool enabled the team to publish successful content, receiving over 500,000 views. Yet creators sometimes treated AI as an unquestioned expert, accepting flawed suggestions. Editorial critique was essential to spot errors and guide creative solutions when AI failed. We discuss how AI's inherent gaps ensure human creativity remains vital.

With the development of AI technologies, especially generative AI (GAI) like ChatGPT, GAI is increasingly assisting people in various tasks. However, people may have different requirements for GAI when using it for different kinds of tasks. For instance, when brainstorming new ideas, people may want GAI to propose different ideas that supplement theirs with different problem-solving perspectives, but for decision-making tasks, they may prefer GAI adopt a similar problem-solving process with people to make a similar or even the same decision as they would. We conducted an online experiment examining how perceived similarities between GAI and human task-solving influence people's intention to use GAI, mediated by trust, for four task types (creativity, planning, intellective, and decision-making tasks). We demonstrate that the effect of similarity on trust (and so intent to use AI) depends on the type of task. This paper contributes to understanding the impact of task types on the relationship between perceived similarity and GAI adoption, with implications for future use of GAI in various task contexts.

With the development of AI technologies, especially generative AI (GAI) like ChatGPT, GAI is increasingly assisting people in various tasks. However, people may have different requirements for GAI when using it for different kinds of tasks. For instance, when brainstorming new ideas, people may want GAI to propose different ideas that supplement theirs with different problem-solving perspectives, but for decision-making tasks, they may prefer GAI adopt a similar problem-solving process with people to make a similar or even the same decision as they would. We conducted an online experiment examining how perceived similarities between GAI and human task-solving influence people's intention to use GAI, mediated by trust, for four task types (creativity, planning, intellective, and decision-making tasks). We demonstrate that the effect of similarity on trust (and so intent to use AI) depends on the type of task. This paper contributes to understanding the impact of task types on the relationship between perceived similarity and GAI adoption, with implications for future use of GAI in various task contexts.

We explore the bi-directional relationship between human and machine learning in citizen science. Theoretically, the study draws on the zone of proximal development (ZPD) concept, which allows us to describe AI augmentation of human learning, human augmentation of machine learning, and how tasks can be designed to facilitate co-learning. The study takes a design-science approach to explore the design, deployment, and evaluations of the Gravity Spy citizen science project. The findings highlight the challenges and opportunities of co-learning, where both humans and machines contribute to each other’s learning and capabilities. The study takes its point of departure in the literature on co-learning and develops a framework for designing projects where humans and machines mutually enhance each other’s learning. The research contributes to the existing literature by developing a dynamic approach to human-AI augmentation, by emphasizing that the ZPD supports ongoing learning for volunteers and keeps machine learning aligned with evolving data. The approach offers potential benefits for project scalability, participant engagement, and automation considerations while acknowledging the importance of tutorials, community access, and expert involvement in supporting learning.

Software projects rely on what we call project archetypes, i.e., pre-existing mental images of how projects work. They guide distribution of responsibilities, planning, or expectations. However, with the technological progress, project archetypes may become outdated, ineffective, or counterproductive by impeding more adequate approaches. Understanding archetypes of software development projects is core to leverage their potential. The development of applications using machine learning and artificial intelligence provides a context in which existing archetypes might outdate and need to be questioned, adapted, or replaced. We analyzed 36 interviews from 21 projects between IBM Watson and client companies and identified four project archetypes members initially used to understand the projects. We then derive a new project archetype, cognitive computing project, from the interviews. It can inform future development projects based on AI-development platforms. Project leaders should proactively manage project archetypes while researchers should investigate what guides initial understandings of software projects.