A New Collaboration Between the Hasso Plattner Institut and HAI Brings the Human Factor of AI to the Forefront

Stanford computer science PhD student Lindsay Popowski has spent the past three years studying social computing, or systems in which human interactions are determined by algorithmic choices. Her work reimagines our online spaces — whether we’re chatting, sharing, or dating — for better social and psychological outcomes. When the invitation came to participate in a new joint research program between the Hasso Plattner Institut (HPI) in Potsdam, Germany, and the Stanford Institute for Human-Centered AI (HAI), she was all in.

“This was an exciting prospect to work and live in a different country and interact with people outside my lab, even outside my normal area of research, for a prolonged amount of time,” Popowski says. She is spending the summer on campus at the Universität Potsdam to work on experimental projects with HPI students and professors. In the fall, she’ll return to Stanford along with a Potsdam student to continue the in-person collaboration with faculty and program leads in the Stanford Human-Computer Interaction Group. 

A Commitment to Diverse Perspectives

The Program on Artificial Intelligence and Human Computer Interaction launched this spring, building on a prior 14-year joint research program between the two universities and their respective design programs. The new program pairs human computer interaction (HCI) and AI students and faculty from both institutes to work on five core research areas: explainability, social computing system design, AI in fabrication, AI-assisted communication, and privacy-preserving AI smart tools. 

“Breakthroughs in science happen at the intersection of fields and when people work together closely,” says HPI Professor Ralf Herbrich, who chairs the institute’s Artificial Intelligence and Sustainability research cluster. "This new program is designed on exactly that, with active exchanges between PhD students at both institutes, co-supervision, and always bringing two diverse points of view into each project. Both institutes will regularly get new perspectives on their research problems through the student exchanges.”

“So far, we’ve seen a heavy focus on the algorithmic side of AI, with the actual interaction of humans with these smart systems as an afterthought. In this program, we are attempting to bring together world experts on both AI and HCI to foster research into human-centered AI systems,” says Professor James Landay, Stanford HAI co-director.

Foundational and Applied Research 

The new collaboration between HPI and HAI will support two foundational research projects and three applied research projects. The two foundational research themes are the following.

Objective Functions for Human-Understandable Explainability: While modern methods of machine learning have reached human-level accuracy, it’s not always possible to explain how a prediction or decision was made to the person affected by it. A team led by HPI’s Herbrich and Landay from HAI proposes to develop objective, machine-optimizable criteria for predictions and decisions, making them understandable to humans. 

Design of Social Computing Systems: The project that PhD exchange student Popowski and colleagues are tackling explores three of the biggest challenges related to our reliance on social networks and other systems that moderate interactions among individuals and groups of people. 

• How can we assess the quality and bias of such services before they are launched in the real world? 

• How can we identify non-human participants in a social network?

• How can we design social algorithms that make intelligent trade-offs among values, goals, and objectives?

“Focusing the research around human-centered AI is an exciting opportunity to push on frontiers that are right at the front of our attention,” says Stanford Associate Professor of Computer Science Michael Bernstein, who will helm this focus area, along with Herbrich from HPI.

The three applied research projects designated for program support cover wide-ranging topics of digital fabrication, AI-based health coaching, and designing smarter home spaces.

AI-Assisted Fabrication Tools: Patrick Baudisch, professor of computer science at HPI and chair of the Human Computer Interaction Lab, is a faculty lead, along with Stanford Professor of Computer Science Maneesh Agrawala, for this effort to use AI to accelerate the process of creating personalized physical objects. The scholars recognize that fabrication tools such as 3D printers, laser cutters, machine tools, and sewing machines have untapped potential for rapid prototyping, but widespread adoption of these tools is hindered by challenges related to tedious calibration, difficult learnability for humans, incompatibilities with common drawing applications, and lack of automation for material prep, assembling parts, and adjusting machine settings.

Baudisch says, “For this project, we’re excited to have HPI digital fabrication researcher Martin Taraz share his approach to calibrating laser cutters and get new inspiration from our HAI counterparts. I expect most of the value of this program will come from the exchange of students.”

AI-Assisted Human-Human Interaction: In domains like mental health and education, many needs go unmet because of the limited availability of human experts. This team jointly led by three faculty — Gerard de Melo, who leads the AI and Intelligent Systems group at HPI; Monica Lam, professor of computer science and faculty affiliate at HAI; and Landay from HAI — sees an opportunity to develop socially intelligent virtual assistants that can hold conversations and show empathy to form meaningful long-term relationships with their users.

AI-Assisted Privacy — Preserving Smart Homes/Offices/Health Spaces: 

Simple controls for regulating light and temperature in homes and offices cause energy waste, if left on when people are not present. AI systems can learn usage patterns in these spaces to help predict when light and climate adjustments are needed. But if individuals are going to trust intelligent sensing systems, these AI tools must maintain privacy. This project, jointly led by Herbrich and Landay, focuses on creating intelligent sensing systems that maintain private information safely, in ways that users will trust. 

Stanford PhD candidate Matthew Joerke has joined the team and additionally hopes to explore ways of using LLM-based health coaching for motivating physical activity. “Being a German-American and working on human-AI interaction in my research, it felt like a perfect fit to get involved. So far, I love living in Berlin/Potsdam and brainstorming new ideas with colleagues and collaborators at the HPI,” he says. 

Besides the student exchanges and ongoing research projects, the program calls for an annual workshop across all major project areas with community-building events to foster deeper connections. 

“We're excited to push the research frontiers of human-centered AI — and grow ourselves and a new generation of researchers along the way,” Bernstein says.

Stanford HAI’s mission is to advance AI research, education, policy and practice to improve the human condition. Learn more.