diff --git "a/out/index.html" "b/out/index.html" new file mode 100644--- /dev/null +++ "b/out/index.html" @@ -0,0 +1,197 @@ +DAT Creativity Game Progress

DAT Creativity Game Progress

An interactive visualization of my journey through the Divergent Association Task (DAT) creativity game, showcasing the hillclimbing optimization technique.

Performance Statistics
Top 0.02%
95.08
Final Creativity Score
Percentile
99.98%
Better than others
Duration
106m
Total game time
Optimization
13
Optimization rounds
Changes
45
Total word swaps
Vocabulary
30
Unique words used

Semantic Distance Analysis

Average Distance
95.1
Across all word pairs
Maximum Distance
104
Most distant pair
Minimum Distance
79
Closest pair (non-zero)

Key Insights

  • • Achieved exceptional creativity score through systematic optimization
  • • Hillclimbing algorithm effectively explored the semantic space
  • 3.46 average word changes per round shows focused refinement
  • • High semantic distances indicate successful divergent thinking
Game Progress Timeline
1

Round 1

11:53 PM
brilliance
sports
peace
rocket
icecream
+5 more
2

Round 2

12:00 AM
stochastic
fuselage
pitfall
screwdriver
sand
+5 more
+10 added-10 removed
3

Round 3

12:00 AM
stochastic
fuselage
pitfall
screwdriver
sand
+5 more
4

Round 4

01:33 AM
stochastic
fuselage
throne
icecream
cricket
+5 more
+6 added-6 removed
5

Round 5

01:33 AM
stochastic
fuselage
throne
icecream
cricket
+5 more
6

Round 6

01:34 AM
stochastic
fuselage
throne
icecream
cricket
+5 more
7

Round 7

01:35 AM
stochastic
fuselage
throne
icecream
cricket
+5 more
8

Round 8

01:36 AM
stochastic
fuselage
throne
icecream
cricket
+5 more
+2 added-2 removed
9

Round 9

01:36 AM
stochastic
fuselage
throne
icecream
cricket
+5 more
10

Round 10

01:38 AM
stochastic
fuselage
throne
icecream
foosball
+5 more
+3 added-3 removed
11

Round 11

01:38 AM
stochastic
fuselage
throne
icecream
foosball
+5 more
12

Round 12

01:39 AM
stochastic
fuselage
throne
icecream
foosball
+5 more
13

Round 13

01:39 AM
stochastic
fuselage
throne
corporate
arthritis
+2 more
-3 removed
Word Evolution & Optimization
Track how words were systematically replaced to maximize semantic distance using hillclimbing optimization.
Round 2
Optimization step
Removed(Lower semantic distance)
brilliance
sports
peace
rocket
icecream
blush
friends
rhythm
velvet
ocean
Added(Higher semantic distance)
stochastic
fuselage
pitfall
screwdriver
sand
spices
arthritis
printer
flute
dinosaurs
This change increased the overall semantic distance between word pairs.
Round 4
Optimization step
Removed(Lower semantic distance)
pitfall
screwdriver
sand
spices
flute
dinosaurs
Added(Higher semantic distance)
throne
icecream
cricket
remote
eel
speedbreaker
This change increased the overall semantic distance between word pairs.
Round 8
Optimization step
Removed(Lower semantic distance)
remote
printer
Added(Higher semantic distance)
corporate
butter
This change increased the overall semantic distance between word pairs.
Round 10
Optimization step
Removed(Lower semantic distance)
cricket
butter
eel
Added(Higher semantic distance)
foosball
chilli
lens
This change increased the overall semantic distance between word pairs.
Round 13
Optimization step
Removed(Lower semantic distance)
icecream
foosball
speedbreaker
This change increased the overall semantic distance between word pairs.

Hillclimbing Strategy

Each word replacement represents a “hill climbing” step where words with lower semantic distances to other words in the set were systematically replaced with words that had higher distances, improving the overall creativity score.

Semantic Distance Matrix

Interactive matrix showing semantic distances between final word pairs

Score: 95.08
stochastic
fuselage
throne
corporate
arthritis
chilli
lens
stochastic
0
103
100
93
94
98
94
fuselage
103
0
94
98
96
91
79
throne
100
94
0
95
98
101
98
corporate
93
98
95
0
97
100
92
arthritis
94
96
98
97
0
86
87
chilli
98
91
101
100
86
0
104
lens
94
79
98
92
87
104
0
Low distance
Medium distance
High distance
Matrix is symmetric

Matrix Analysis

Average Distance:
95.1
Highest Distance:
104
Lowest Distance:
79
Understanding Hillclimbing Optimization

Hillclimbing is a local search optimization algorithm that continuously moves towards better solutions by making incremental improvements. Think of it like climbing a hill in foggy weather - you can only see your immediate surroundings, so you always take the step that goes uphill.

Algorithm Steps

1
Initialize: Start with a random solution (10 diverse words)
2
Evaluate: Calculate semantic distances between all word pairs
3
Find weakness: Identify the word pair with smallest distance
4
Improve: Replace one word with a more distant alternative
5
Repeat: Continue until no improvement is possible

Application to DAT

Objective Function

Maximize the sum of semantic distances between all word pairs

Neighbor Generation

Replace one word at a time with alternatives from vocabulary

Local Optimum

Stop when no single word replacement improves the score

Strategy Insights

Why Hillclimbing Works Here

  • Greedy improvement: Each word replacement immediately improves the score
  • Fast convergence: Reaches a good solution quickly
  • Intuitive process: Mimics how humans naturally optimize

Potential Limitations

  • Local optima: May get stuck in suboptimal solutions
  • Initial dependency: Result depends on starting words
  • Single-word changes: Cannot make coordinated multi-word swaps

My Implementation Result

Despite these theoretical limitations, the hillclimbing approach was highly effective for the DAT creativity game, achieving a score of 95.08 and ranking in the 99.98th percentile.

This success suggests that the semantic distance landscape for this problem has favorable properties for local search, with relatively few problematic local optima.

What is the DAT Creativity Game?
Understanding the Divergent Association Task

The Divergent Association Task (DAT) is a creativity assessment that measures your ability to think of diverse, unrelated concepts. Players must generate 10 words that are as semantically distant from each other as possible.

The challenge lies in finding words that share minimal conceptual overlap, requiring creative thinking and broad knowledge across different domains.

\ No newline at end of file