Large scale ideation has developed as a promising new way of obtaining large numbers of highly diverse ideas for a given challenge. However, due to the scale of these challenges, algorithmic support based on a computational understanding of the ideas is a crucial component in these systems. One promising solution is the use of knowledge graphs to provide meaning. A significant obstacle lies in word-sense disambiguation, which cannot be solved by automatic approaches. In previous work, we introduce \textit{Interactive Concept Validation} (ICV) as an approach that enables ideators to disambiguate terms used in their ideas. To test the impact of different ways of representing concepts (should we show images of concepts, or only explanatory texts), we conducted experiments comparing three representations. The results show that while the impact on ideation metrics was marginal, time/click effort was lowest in the images only condition, while data quality was highest in the both condition.
