In Practice Notes, you’ll hear from Local Center teams and our own staff about what we’re learning as we support community-led design projects across New York City.

By Janice Chu

Through our work with Local Center, di Domenico + Partners partnered with Sunnyside Shines BID to help develop a vision plan for the community’s visions for their public space. The Sunnyside Vision Plan is an actionable roadmap for enhancing public spaces, strengthening community life, and supporting neighborhood growth. As both a designer and an outreacher in the project, integrating community input into design is a constant process, learning how to listen has become an important question. Among all our engagement efforts to shape the Sunnyside Vision Plan, the workshop at the Aviation High School was a unique one — one where we tested new ways of surfacing ideas early in the design process. We started designing the workshop by rethinking traditional engagement strategies: How can we better listen to and design with a younger generation group? Beyond traditional analog tools, like post-its, one-on-one conversations or surveys, can we improve the flow of conversation, gather richer input, and better visualize community ideas to support deeper engagement?

These questions led us to experiment with Generative AI as part of a co-design workshop. While the design field has been cautious in adopting AI — concerned it might dilute empathy and place-based values — we approached it as a support tool to inspire conversation, increase efficiency in gathering and analyzing ideas, and create a visual catalog for designers to reference directly.

Step 1: Contextualizing

We started the workshop by asking everyone “How would you describe Sunnyside?” In a collective brainstorming session, participants shared what makes the neighborhood unique and reimagined a familiar underused space: the stretch beneath the 7 train. As a space that everyone lives near or passes by every day, participants were eager to share personal stories, memories, and ideas for improvement. Comments were live-updated on the screen, encouraging a shared, multi-perspective discussion that builds on each other’s thoughts.

Interactive presentation with live-updated comments that sparks conversations.

Step 2: dreaming

We then guided participants to fill out a curated worksheet: what, who, where, and how people and objects occupy the space in their imagination. Once completed, the worksheets were fed into a pre-trained AI model that turned the text into vivid images, bringing to life ideas that previously remained unexplored. We invited participants to respond to the printed images by commenting, voting, and comparing them to their original visions.

Participants’ comments on AI generated design iterations.

Step 3: city making

The workshop concluded with a collage session, where participants cut out elements they resonated with from dozens of images to assemble their final vision. Most final collages drew ideas from others’ imaginations — for example, one design brought together sports, art, and small businesses beneath the bridge, flanked by greenery, expressing a desire for a multi-use public space welcoming to all ages.

The workshop demonstrated that when thoughtfully curated, Generative AI can amplify community voices, not replace them.

Worksheets filled out by participants, AI generated design iterations and Individual Collages.

Embedding Inclusivity, Empathy, and Sense of Place into Generative AI for Engagement

The accessibility of Generative AI made the session intuitive, allowing participants to create detailed visuals regardless of design background. The AI-generated images provided a shared visual language between designers and participants, enabling quicker feedback and more dynamic conversations. With live-updated comments during the presentation, we adjusted engagement strategies in real time and effectively gained insight into how people imagine their environment.

Participants reviewing AI generated images, and exchanging thoughts with the project team.

Acknowledging the limitations and biases in Generative AI models — often rooted in internet-sourced training data that defaults to dominant cultural norms and commercial aesthetics — we implemented strategies to avoid producing generic or irrelevant content.

To emphasize local context, we curated a worksheet that grounded each prompt in a familiar element: the concrete arches beneath the 7 train viaduct in Sunnyside. By setting a consistent baseline, image generations maintained contextual relevance while design outcomes varied based on each participant’s input.

We guided participants in crafting detailed prompts based on their own experiences and preferences, encouraging the use of adjectives and descriptions to better constrain the AI model. The worksheet asked them to imagine all aspects of the space, from overall activity and user groups to focal objects and redesigning the ground or ceiling. Structuring the input helped the AI model to generate more specific, meaningful results.

Generating multiple variations of each prompt provided a range of visual options for collage-making, adding design liberty to the process. Instead of working with one predesigned image, participants sorted through dozens to find the playground, mural, or lighting element that best matched their vision.

In addition to digital tools, participants also used analog means — post-its, sketches, notes, and cut-and-paste collages — to express ideas. These hands-on methods allowed for nuanced storytelling and socio-cultural values that AI alone could not capture. Each final vision was a layered, multimedia collage, injected with personal thoughts and aesthetic. Together, these strategies complemented each other: technology offered speed and breadth, while human touch brought depth and meaning.

Participants collage images to envision their final reimagination of the space

Reflections and Moving Forward

As part of the many engagement efforts for the Sunnyside Vision Plan, this workshop was our first attempt at exploring technology-supported co-design strategies with the public during the early visioning stage. It surfaced values and perspectives we hadn’t encountered in other engagement sessions, and fostered a more immediate, aligned dialogue between us as designers and the community.

With a group of young participants comfortable with digital tools, the process was enjoyable and engaging. We were surprised by the depth of creativity once participants were given the right tools — many ideas went beyond what we imagined. Specific desires, like having sports courts below the 7 train, hadn’t come up in previous sessions with other engagement groups. The workshop also challenged the traditional engagement format, where public input is gathered and revisited months later, often without clear share-back. Instead, this session allowed for real-time visualization, feedback, and iteration. With the right setup, this model could evolve into a more interactive format and be adapted to different age groups and project types moving forward.

For a typical engagement session, designers define the questions, guide the feedback, and interpret local ideas into design. While this is what we are trained for, we’re learning there’s a fine line between facilitating dialogue and oversteering the agenda. In our experiment with Generative AI, the tool acted as a third voice in the design conversation: a direct channel that was guided and constrained by community input, to interpret and generate people’s imaginations directly. This helped shift some creative authorship to the community. 

We left the workshop with a rich set of collages, notes, and drawings:a living archive of public imagination. Each collage traced a design idea from conversation to AI-generated image to a unique photomontage, grounded in everyday experiences and cultural nuance. As this methodology continues to evolve, we see its potential to deepen and accelerate community ideas, and to build more open, responsive co-design processes.

Janice Chu is an architectural designer at di Domenico + Partners and a member of the Sunnyside Local Center team.

Image Credit: Courtesy of dD+P