Human-AI
Interaction Lab
Our Mission
Building more just and empowering workplaces and cities by creating technology that supports and strengthens individual and collective human decision-making.
Highlights
Taken for a Ride: The Hidden Costs of Algorithmic Management for Rideshare Workers
We’re excited to share our website Demystifying Gigwork that distills findings from several of our research papers into one readable, public-friendly interface! It’s meant for anyone, no matter your knowledge about gig work. Come explore, learn more, and try out one of our data tools we created to see how much you’d earn as a rideshare driver! Let us know how you’re getting on with it through our feedback form at the bottom. We look forward to hearing for you!
Algorithmic management and occupational health: A comparative case study of organizational practices in logistics
Algorithmic Management (AM), which refers to technologies that use algorithms to oversee and direct workers, is increasingly being introduced across various sectors and workplaces. While previous research has focused on AM's impact on job quality in platform work, its effects on worker well-being in non-platform workplaces remain underexplored. This study seeks to deepen our understanding of the impact of AM on occupational health within non-platform workplaces. Drawing on the socio-technical lens and the Pressure, Disorganization and Regulatory Failure (PDR) model (Quinlan et al., 2001), it aims to identify organizational practices that shape the interplay between AM and employees' work experiences, health, and overall well-being. We conducted a comparative case study with two Swedish logistics companies and collected data from observations and semi-structured interviews. Our analysis focused on the interplay between organizational practices, AM technology, and worker experiences to understand key differences between the cases. Workers at both sites reported a low sense of autonomy and task significance. However, physical and psychological strain from AM was more pronounced in the e-commerce company, a disparity potentially explained by factors of the PDR model. We identified organizational practices that appear to positively influence workers' AM experiences: i) involving workers with the AM technology; ii) integrating AM considerations into occupational safety and health management; iii) designing AM applications that allow worker control; and iv) managerial practices that add qualitative assessments to AM's quantitative evaluations. Our research highlights the critical importance of designing organizational practices that incorporate AM in ways that promote occupational health alongside operational efficiency.
Proxona: Supporting Creators' Sensemaking and Ideation with LLM-Powered Audience Personas
A content creator’s success depends on understanding their audience, but existing tools fail to provide in-depth insights and actionable feedback necessary for effectively targeting their audience. We present Proxona, an LLM-powered system that transforms static audience comments into interactive, multi-dimensional personas, allowing creators to engage with them to gain insights, gather simulated feedback, and refine content. Proxona distills audience traits from comments, into dimensions (categories) and values (attributes), then clusters them into interactive personas representing audience segments. Technical evaluations show that Proxona generates diverse dimensions and values, enabling the creation of personas that sufficiently reflect the audience and support data-grounded conversation. User evaluation with 11 creators confirmed that Proxona helped creators discover hidden audiences, gain persona-informed insights on early-stage content, and allowed them to confidently employ strategies when iteratively creating storylines. Proxona introduces a novel creator-audience interaction framework and fosters a persona-driven, co-creative process.
Current Projects
Reimagining Algorithmic Management for Worker Well-Being through Worker Co-Design and Policy Approaches
Increasingly, people are turning to gig work platforms for flexible work opportunities, yet paradoxically, research has shown that algorithmic management controls gig workers through tactics such as gamified incentives and opaque work assignment and commission rates. We use co-design methods to engage with stakeholders—e.g., workers, organizers, and policymakers—to understand how algorithmic management impacts worker well-being, surface ideas for interventions to support worker protections, and design tools to help with policymaking. We are currently designing policymaking training sessions to illustrate to policymakers what algorithmic management is and how it impacts workers and to surface policy needs—e.g., specific wording to use in bills, and data needed to help garner colleague support. We are also collaborating with partners at CMU and UMN to prototype a data-sharing system for gig workers and policymakers to investigate worker issues and inform related policies.
Designing Tools to Support Organizational Decision-Making and Support Participatory AI Design
One concern of AI technologies is its potential to generate inequitable outcomes for underrepresented populations, often linked to overlooked historical human biases within datasets or lack of input from impacted constituents. This has led to increased calls eschewing AI automation for AI assistance. Yet, AI assistance still encompasses challenges including 1) humans drawing on AI assistance can still exhibit biased decision-making, and 2) in organizations, it can be unclear how to resolve differences in what values to embed in AI assistive tools. To that end, we explored how to support organizational deliberation to support more inclusive outcomes through focus groups where participants used historical data to construct personal models about Master’s admissions and surface patterns of organizational decision-making to inform future practices. Based on how participants used their personal ML models as boundary objects to deliberate together and share situational contexts informing their preferences, we are currently exploring how to systematically integrate impacted stakeholder participation in the data exploration phase of AI design for whether this can surface diverse perspectives for (un)acceptable uses of data.
News and Announcements