Galton's Law of Mediocrity: Why Large Language Models Regress to the Mean and Fail at Creativity in Advertising

TL;DR

The paper investigates whether large language models truly exhibit creativity or regress toward mediocrity in advertising tasks. It introduces Galton’s Law of Mediocrity to describe a bias in next-token prediction that privileges common patterns over novelty. A two-phase creativity stress test—Phase 1 forgetting and Phase 2 expansion—applies to 1,045 ad concepts across multiple LLMs, using both plain and marker-driven expansion alongside quantitative metrics (cosine similarity, METEOR, entropy, 4-gram uniqueness) and qualitative judgments. Findings show that forgetting erases creative content first, expansions generate surface-level novelty but fail to restore original depth, and marker-guided expansion improves alignment and diversity but cannot fully recover originality. The work highlights the need for creativity-aware objectives and prompting strategies to develop LLMs that sustain genuine originality in creative tasks like advertising.

Abstract

Large language models (LLMs) generate fluent text yet often default to safe, generic phrasing, raising doubts about their ability to handle creativity. We formalize this tendency as a Galton-style regression to the mean in language and evaluate it using a creativity stress test in advertising concepts. When ad ideas were simplified step by step, creative features such as metaphors, emotions, and visual cues disappeared early, while factual content remained, showing that models favor high-probability information. When asked to regenerate from simplified inputs, models produced longer outputs with lexical variety but failed to recover the depth and distinctiveness of the originals. We combined quantitative comparisons with qualitative analysis, which revealed that the regenerated texts often appeared novel but lacked true originality. Providing ad-specific cues such as metaphors, emotional hooks and visual markers improved alignment and stylistic balance, though outputs still relied on familiar tropes. Taken together, the findings show that without targeted guidance, LLMs drift towards mediocrity in creative tasks; structured signals can partially counter this tendency and point towards pathways for developing creativity-sensitive models.

Figures (19)

• Figure 1: Phase 1 Forgetting Trends Across Models
• Figure 2: Marker-Driven: Original advertisement with extracted markers (bottom).
• Figure 3: Comparison of marker performance.
• Figure 4: Phase-2 trends across Model.
• Figure 5: Phase-2 Marker-Driven trends across levels.