In partnership with

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

Sign up now for free and work smarter, not harder.

Navigating Subjectivity in Design

Hello Fellow Magicians,

Last week, we explored computation vs. AI in AEC, breaking down why computation excels in rule-based tasks but struggles in ambiguous, subjective scenarios. This week, I want to dive deeper into how AI helps where computation falls short.

If computation is about strict rules, AI is about understanding the gray areas—the places where human intuition, preferences, and evolving project needs make strict logic impossible to define upfront.

There are two major advantages of AI systems over computational systems:

1) Encoding Subjectivity – AI can learn from patterns in past decisions, enabling it to navigate design choices where there is no single “right” answer.

2) Handling Multiple Reasonable Outputs – Instead of enforcing a single optimal solution, AI can generate a range of valid possibilities, allowing architects and engineers to explore different trade-offs.

In the sections ahead, we’ll explore how AI enables more flexible, context-aware decision-making—turning it into a powerful tool for AEC professionals dealing with complexity, ambiguity, and evolving constraints.

Encoding Subjectivity of Design

Architecture and design are rarely about rigid, predefined answers. Unlike engineering calculations or rule-based automation, many design decisions are subjective—driven by aesthetics, experience, and intuition rather than fixed mathematical formulas.

Computation struggles here because it depends on clearly defined inputs and outputs. It requires explicit rules like:

• "This wall must be exactly 12 feet long."

• "This column must support X pounds of load."

• "This HVAC system must provide Y cubic feet per minute of airflow."

But what about decisions that can’t be boiled down to simple rules?

• What makes a space feel open and welcoming?

• What’s the best way to arrange spaces for collaboration vs. privacy?

• How should a façade’s geometry respond to both function and aesthetic preference?

These are gray areas where different architects may have different opinions, and even the same person might change their mind based on context.

This is where AI excels. Instead of relying on explicit rules, AI can learn from past design patterns, user preferences, and project data to make informed, context-aware suggestions.

Learning Complicated Design Patterns

AI approaches subjectivity very differently from computation. Instead of relying on sets of explicit rules, AI systems learn by identifying patterns and recognizing relationships. This allows AI to handle gray areas in design—where rules are hard to define the best solution.

So how does AI learn complicated design patterns from subjective and ambiguous data? Let’s explore:

Training on Large, Varied Datasets

Contrasting computation which starts with defining useful rules - developing a machine learning system starts with training a machine learning model on existing data that represents the task you would like it to learn. You might decide to teach it through:

• Previous architectural designs (BIM models, floor plans, and spatial layouts).

• User feedback (e.g., what architects typically revise in a design).

• Environmental and contextual data (sunlight patterns, circulation flow, adjacency preferences).

Example: An AI system analyzing thousands of office layouts can learn how different workspace configurations balance privacy, collaboration, and circulation—without needing hardcoded rules.

Key Mechanism: AI detects recurring spatial patterns and relationships that emerge across projects. Instead of saying, "All open workspaces must have walls exactly 8 feet high," AI can recognize when lower walls improve collaboration and when enclosed spaces enhance productivity. Combined with an understanding of what types of rooms require productivity and which require collaboration - the system can learn to apply the design rules accordingly.

Identifying Latent Features in Design

A big advantage of AI is its ability to recognize latent features— relationships in data that humans might not explicitly define.

Example: Architects designing facades might focus on:

• Aesthetic consistency (alignment of windows, symmetry, or materiality).

• Performance factors (solar gain, ventilation efficiency).

• Contextual fit (relationship to surrounding buildings).

While traditional computation requires explicit instructions (e.g., “Make sure all windows align”), AI can infer design principles by analyzing thousands of successful projects.

Key Mechanism: AI doesn't just copy past designs—it learns abstract concepts like proportion, rhythm, and adjacency, enabling it to suggest new designs that align with human preferences without needing strict formulas.

Refining Outputs Through Human Feedback

Unlike computation, which produces rigid, rule-based outputs, AI can continuously improve through human feedback loops.

Example: An AI system generates a series of conceptual layouts for a new building.

• The architect selects the most promising option.

• They make manual refinements to adjust circulation flow.

• The AI learns from these modifications and incorporates them into future suggestions.

Key Mechanism: AI doesn’t just optimize based on historical data—it adapts to real-time user input, improving over time.

NOTE:

Developing AI systems that automatically learn via human feedback is not as simple as it sounds - some systems do this - typically they are called “Human in the Loop” systems. I’ll discuss these further in a later field note

Managing the Gray

By leveraging pattern recognition, learning from feedback, and balancing competing priorities, AI steps in where traditional computation falls short—handling gray-area design challenges that can't be reduced to strict rules. Instead of enforcing a single solution, AI enables adaptive, context-aware decision-making, allowing architects and engineers to explore multiple viable options.

🚀 Next week, we’ll dive deeper into how AI systems can be designed to generate and manage multiple reasonable outputs, expanding creative possibilities in AEC. Stay tuned!