GameNet and GameSage: Videogame Discovery as Design Insight

GameNet and GameSage present a videogame discovery system designed to help game designers locate related prior work in an industry where the volume of new releases has outpaced any individual's ability to survey existing games. The authors frame this discoverability problem as a contributor to blockbuster-driven, bandwagon-style development, and propose a network-based approach as a remedy. GameNet is a network of nearly 12,000 linked games built using latent semantic analysis trained on Wikipedia articles, while GameSage accepts a free-text game idea and returns the most closely related existing titles. The work is positioned as the first formal, controlled user study of a tool of this kind.

The evaluation involved 182 undergraduate game-design students who, over a 15-minute session, searched for games related to their own projects, using either the dedicated tools or general web search. Participants using GameSage and GameNet discovered substantially more games per person, averaging 14.16 compared with 6.49 for the baseline, and a far higher proportion of those games were unfamiliar to them, at 80.3% versus 47.0%. The difference in the uniqueness of discovered games was not statistically significant, so the diversity hypothesis was not supported, though the novelty finding was strong.

The authors temper these results with important constraints. The study used a fixed condition order and a small, non-expert sample, which limits generalization to professional designers. Only 49% of participants indicated they were at least somewhat likely to reuse the tools, and this preference did not correlate significantly with query clarity. The underlying Wikipedia-trained model underrepresents novel games and introduces extraneous non-design content, both acknowledged as limitations. Following public release, the system logged 7,399 sessions from 6,488 users, indicating meaningful real-world uptake beyond the controlled experiment.