Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Are Semantic Networks Associated with Idea Originality in Artificial Creativity? A Comparison with Human Agents

Umberto Domanti, Lorenzo Campidelli, Sergio Agnoli, Antonella De Angeli

TL;DR

This study investigates whether semantic memory networks relate to idea originality in artificial creativity and how GPT-4o compares to human agents. By analyzing verbal production tasks and constructing semantic networks via TMFG from both humans (split into higher and lower creative groups) and GPT-4o responses to AUT prompts, the authors test two sub-hypotheses about originality and network flexibility. They find that higher-creative humans are more original and have more flexible, robust semantic networks, while GPT-4o is less flexible but can surpass lower-creative humans in originality; the relationship between network structure and originality is strongest for the human high-creative group. The work provides empirical, operational, and methodological advances for designing and evaluating Creativity Support Tools, highlighting the potential and limits of artificial creativity and offering a research agenda that links cognitive theory, network analysis, and human-computer interaction.

Abstract

The application of generative artificial intelligence in Creativity Support Tools (CSTs) presents the challenge of interfacing two black boxes: the user's mind and the machine engine. According to Artificial Cognition, this challenge involves theories, methods, and constructs developed to study human creativity. Consistently, the paper investigated the relationship between semantic networks organisation and idea originality in Large Language Models. Data was collected by administering a set of standardised tests to ChatGPT-4o and 81 psychology students, divided into higher and lower creative individuals. The expected relationship was confirmed in the comparison between ChatGPT-4o and higher creative humans. However, despite having a more rigid network, ChatGPT-4o emerged as more original than lower creative humans. We attributed this difference to human motivational processes and model hyperparameters, advancing a research agenda for the study of artificial creativity. In conclusion, we illustrate the potential of this construct for designing and evaluating CSTs.

Are Semantic Networks Associated with Idea Originality in Artificial Creativity? A Comparison with Human Agents

TL;DR

This study investigates whether semantic memory networks relate to idea originality in artificial creativity and how GPT-4o compares to human agents. By analyzing verbal production tasks and constructing semantic networks via TMFG from both humans (split into higher and lower creative groups) and GPT-4o responses to AUT prompts, the authors test two sub-hypotheses about originality and network flexibility. They find that higher-creative humans are more original and have more flexible, robust semantic networks, while GPT-4o is less flexible but can surpass lower-creative humans in originality; the relationship between network structure and originality is strongest for the human high-creative group. The work provides empirical, operational, and methodological advances for designing and evaluating Creativity Support Tools, highlighting the potential and limits of artificial creativity and offering a research agenda that links cognitive theory, network analysis, and human-computer interaction.

Abstract

The application of generative artificial intelligence in Creativity Support Tools (CSTs) presents the challenge of interfacing two black boxes: the user's mind and the machine engine. According to Artificial Cognition, this challenge involves theories, methods, and constructs developed to study human creativity. Consistently, the paper investigated the relationship between semantic networks organisation and idea originality in Large Language Models. Data was collected by administering a set of standardised tests to ChatGPT-4o and 81 psychology students, divided into higher and lower creative individuals. The expected relationship was confirmed in the comparison between ChatGPT-4o and higher creative humans. However, despite having a more rigid network, ChatGPT-4o emerged as more original than lower creative humans. We attributed this difference to human motivational processes and model hyperparameters, advancing a research agenda for the study of artificial creativity. In conclusion, we illustrate the potential of this construct for designing and evaluating CSTs.
Paper Structure (64 sections, 4 figures, 7 tables)

This paper contains 64 sections, 4 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. HCI and Creativity
  4. Creativity as a Cognitive Process
  5. Divergent Creativity
  6. Semantic Memory
  7. Method
  8. Participants and Design
  9. Procedure
  10. Experimental Tasks
  11. Semantic Networks
  12. Results
  13. Originality
  14. Total Production
  15. Originality Peaks
  16. ...and 49 more sections

Figures (4)

  • Figure 1: Creative Cognition framework, adapted by benedek_toward_2019
  • Figure 2: Kernel density estimates of mean originality scores across groups
  • Figure 3: 2D representation of GPT, LCH, and HCH semantic networks
  • Figure 4: Percolation analysis. On the left (a): Representation of the number of connected nodes in each network at increasing thresholds (I). On the right (b): Comparison of the percolation integrals between groups