The communication of precautionary information through symbols or pictorials has the potential to overcome language and educational barriers. Language independent communication is particularly important for populations that may not have otherwise received a text-based warning, such as non-English speaking immigrant workers. The primary objective of this research was to develop, implement, and evaluate a novel approach to symbol development utilizing interactive evolutionary computation (IEC) to expand end-users participation in the design process. At present, graphic designers, guided by their own intuition and creativity and usually working in isolation from the intended audience, develop the symbols used in warnings. Symbols are often not adequately evaluated prior to their introduction and may not be well suited for communicating their intended message. The use of the production method, where user input is incorporated in the design process, increases the level of participation and includes subjective preference to a limited extent.

IEC is one computer-based method for increasing the involvement of end-users in design that explores many possible solutions to a given design problem through humancomputer collaboration. Specifically, this class of algorithms exploits an analogy between the design process and biological evolution (i.e., natural selection). A human designer guides the evolutionary process by judging the quality of novel solutions generated by the IEC algorithm, which evolve from one generation to the next based upon the user’s input.

In this study, three participant groups, comprised of students as well as immigrant and non-immigrant industrial workers, used the IEC algorithm to develop 120 symbols conveying precautionary information. The symbols were evenly divided between two distinct referents, a laceration injury and awkward postures during MMH. These symbols were used to form composite symbols using three different methods:​ taking the median angles from stratification based on personal characteristics;​ stratification based on kmeans cluster analysis;​ and stratification based on factor analysis in combination with kmeans clustering. The quality, appropriateness, and comprehensibility of the resulting symbols was judged by the design group as well as a validation participant group. The symbols were evaluated in terms of preference rankings (relative comparisons) and comprehension estimates (independent scores).

The results suggest that IEC algorithms can be successfully integrated into a usercentered process of symbol design. Significant differences in symbol parameter values were found for both referents based on designer familiarity with the hazards. Symbol evaluations were similar in the preference rankings and comprehension estimates of the design and validation subgroups, which implies no carry-over effect (bias) affected the evaluation of candidate symbols. As such, one can conclude that only a sample of the target population is needed to create comprehensive symbols using the IEC algorithm.

The effectiveness of IEC output was also evaluated against symbols generated through traditional graphic design methods. Laceration composite symbols compared favorably to symbols identified by woodworking industrial groups as standard symbols for this hazard. The best performing IEC composite symbols were statistically equivalent to the best performing industry symbols in terms of both preference rankings and comprehension estimates.

This study indicates that workers exposed to a hazard can develop warning symbols that are likely to be well understood by their peers. The clear and effective communication of precautionary information in the workplace is crucial to the prevention of accidents and subsequent injuries. The use of symbols may promote effective communication of these hazards, especially to workers who are not fluent readers of the native language. Actual or potential applications of this research include a new approach to creating symbols that encourages end-user input into their design.