Layer 3. Compare and justify – detailed design preferences

The third layer drills down into preferences at a more detailed level by isolating and examining the component parts that fit together to shape places.

The built environment is constructed from many component parts – roofs, facades, street trees, pavements, and so on. This methodological layer allows participants to compare and express preferences relating to images of those component parts. But because any component can take many forms, a balance needs to be struck between showing enough images to capture this diversity and not overwhelming participants with the sheer variety. The key is careful image selection and encouraging participants to justify their choices.

Step 1: Select components to be investigated

The approach focusses on specific design components and is unsuited to ‘whole scene’ views such as those evaluated in Layer 2 where too many variables are present in every image for meaningful comparisons to be made. Focussing only on components that can be clearly isolated in images and compared without other factors confusing the picture, identify the design components to be investigated. In Urban Beauty, knowing the unknowable these were:
1. Street trees
2. Parking
3. Green infrastructure
4. Social spaces
5. Play spaces
6. Street surfaces
7. Street boundaries
8. Facade variation
9. Facade design
10. Roofscapes
11. Building materials

Step 2: Select types and construct collages

For each component construct one or more collages of images. Nine images in a grid works well for participants to concurrently view and compare images without being overwhelmed. For particularly complex components, such as facade design, more than one collage may be required. But rather than attempting to include every possible variation of a component, identify the main ‘types’ of each and select a representative image to be included in the collage as depicted in the example for threshold situations in urban contexts. When shown to participants, these should be randomly placed in the grid and unlabelled. Like layer 2, they should be context specific, with participants shown collages that relate to the sort of context in which they live.

Step 3: Ensure images are comparable

Again, select images that reflect a broad diversity of situations, housing typologies, architectural styles, viewing points, and (if appropriate) densities, to help participants focus on the design component. Then, digitally edit the images to remove distracting or confounding factors, such as people and obviously different weather conditions. This can easily be done in any standard photo editing software or many generative AI chatbots now have the facility to manipulate images.

Step 4: Ask participants to identify the ‘best’ and ‘worst’ examples

Provide clear instructions for comparing the images, giving participants the task of selecting the best and worst examples from each image. For example: “The focus now shifts to particular aspects of residential design. In each collage click once on the best designed example and click twice on the worst designed example (click a third time to remove a selection).”

Step 5: Invite participants to explain their choices

After selecting the best and worst examples, a critical stage is the opportunity for participants to justify their choices in an open text box. This step is crucial because it captures the nuanced reasoning behind preferences, something that conventional visual preference surveys often fail to do.

Step 6: Analyse the data quantitatively

First analyse the responses using a simple scoring system:
1. Disliked
2. Neutral
3. Liked
Average scores can then be calculated for each image and compared across participant groups. In the example from the Urban Beauty, knowing the unknowable study, these can be seen in relation to perceptions of street trees in urban areas among local councillors (bottom left in each image) and planning officers (top right). Colours help to visualise the results:
• Blue indicates the highest scoring examples
• Green indicates the second highest
• Yellow indicates predominantly neutral responses
• Orange indicates the second lowest
• Red indicates the lowest scoring.

Step 7: Qualitatively identify broader themes and hidden influences

Next analyse the written explanations provided by participants to help identify the factors that shape participant preferences, including the ‘unseen’ considerations that influence visual preferences, such as the functional, ethical, or social issues that people infer from what they see. This involves:
• Coding comments to identify related ideas
• Grouping related responses together
• Using these themes to better interpret the quantitative design component data and rankings.