Designing the Scoring Model

Due: Wednesday, April 3

Moodle

Overview

In this assignment, your group will design the scoring model that defines what makes a “good” gallery wall.

Rather than assuming a fixed definition of quality, your goal is to design a system that represents curatorial preferences in a structured way. This scoring model will later be used by an algorithm to generate and evaluate candidate gallery layouts.

This is the foundation for everything that follows.

Objective

Design a scoring.yaml structure that:

  • captures what the gallery cares about
  • represents tradeoffs (not just “maximize everything”)
  • supports human-guided improvement of layouts
  • can realistically be used by an algorithm

Part 1: Proposed Scoring Model

Create a YAML-style structure with at least 4–6 criteria.

Use the following as a starting template:

criteria:
  - id: example_criterion
    label: "Short human-readable name"
    description: "What does this measure, and why does it matter?"
    scale:
      min: 0
      max: 100
      low_meaning: "What does a LOW value represent?"
      high_meaning: "What does a HIGH value represent?"
    preferred_value: 50
    tolerance: 20
    importance: 0.5

You may modify or extend this structure, but you must explain your design choices.

Example Criterion (Concrete)

To make this more concrete, here is one fully worked example of a criterion.

- id: thematic_cohesion
  label: "Theme cohesion vs diversity"
  description: "Measures how similar the artworks are in subject or theme."
  scale:
    min: 0
    max: 100
    low_meaning: "artworks are very different in theme (high contrast)"
    high_meaning: "artworks share a strong common theme"
  preferred_value: 70
  tolerance: 15
  importance: 0.8

How to Think About This Example

  • A value near 0 means the wall contains very different kinds of artworks
  • A value near 100 means the artworks are highly related in theme

The curator is saying:

  • “I want fairly strong thematic unity” (preferred value = 70)
  • “But it doesn’t need to be perfect” (tolerance = 15)

So:

  • A wall with value 65–85 would likely be considered good
  • A wall with value 20 (too random) or 95 (too uniform) would score worse

Key Insight

This is not a “maximize this” metric.

It is a target-based metric:

The goal is to get close to the preferred value, not to push to the extreme.

Important

You should not just copy this example.

Your task is to:

  • decide what criteria matter
  • define what each scale means
  • determine appropriate preferred values
  • justify your design choices

Example Criterion (Arrangement-Based)

This example focuses on how artworks are positioned relative to each other, rather than just which ones are selected.

- id: adjacency_contrast
  label: "Adjacency contrast"
  description: "Measures how different neighboring artworks are in style, color, or theme."
  scale:
    min: 0
    max: 100
    low_meaning: "neighboring artworks are very similar"
    high_meaning: "neighboring artworks are highly contrasting"
  preferred_value: 50
  tolerance: 20
  importance: 0.7

How to Think About This Example

This criterion is about local relationships on the wall:

  • Do adjacent artworks feel repetitive?
  • Or do they clash too strongly?

The curator is saying:

  • “I want some contrast, but not too much” (preferred value = 50)
  • “A moderate range is acceptable” (tolerance = 20)

So:

  • A wall with value 30–70 would likely be good
  • A wall with value 10 (too repetitive) or 90 (too chaotic) would score worse

Why This Example Matters

This is different from the previous example in an important way:

  • Thematic cohesion depends mostly on which artworks are chosen
  • Adjacency contrast depends on how those artworks are arranged

This highlights a key idea:

Some criteria evaluate the set of artworks, while others evaluate their placement and relationships.

Key Insight

Again, this is not about maximizing contrast.

It is about finding a balanced level that fits the curator’s intent.

Part 2: Explanation of Criteria

For each criterion, explain:

  • Why did you include it?
  • Does it measure:
    • selection (which artworks are chosen),
    • arrangement (how they are placed), or
    • both?
  • Should this be:
    • maximized,
    • minimized,
    • or matched to a target value?

Part 3: Opposing Dimensions

Identify at least two criteria that represent a spectrum, such as:

  • unity vs diversity
  • harmony vs contrast
  • consistency vs surprise

For each:

  • Explain why it should be treated as a target range
  • Describe what happens if the system pushes too far in either direction

Part 4: Scoring Strategy

Describe how your system evaluates a wall:

  • How do you compare actual values to preferred values?
  • How does tolerance affect scoring?
  • How are multiple criteria combined into a final score?

Be specific enough that someone could implement your approach.

Part 5: Reflection

Briefly answer:

  • What was the hardest part of designing this?
  • What assumptions are you making about the curator?
  • Where might your model fail?

Deliverable

Submit a single document (PDF or Markdown) that includes:

  • your scoring.yaml proposal
  • explanations for all sections
  • reflection

Be clear, structured, and concise.

Scroll to Top