Foundations for an Empirically Determined Scale of Trust in Automated Systems

Abstract Hazard communications often contain adverbial or adjectival phrases that qualify the meaning of verbs or nouns. For example, a warning label might contain the phrase prolonged exposure. Fuzzy set theory provides a method of quantitatively describing the meaning of such phrases, using membership functions. For example, a membership function might map how strongly workers feel exposure times between 0 and 24 hr belong to the fuzzy set prolonged exposure. The focus of this study was on documenting worker understanding of adverbial or adjectival phrases used in hazard communications using fuzzy sets. Emphasis was placed on documenting how membership functions changed between contexts. To do this, 353 workers at 9 locations were asked to interpret the meaning of hazard communication phrases containing fuzzy qualifiers, such as "safe," "weak," "strong," "prolonged," "extremely," "slightly," "poorly," or "adequate." Workers interpreted the phrases in meaningful ways, confirming that fuzzy qualifiers can play a useful role in hazard communication. Terms, on opposite poles of constructs, such as "extremely versus slightly" or "weak versus strong," shifted the obtained membership functions consistently with predictions. Including contextual information within the hazard communication phrases also reduced worker uncertainty, and resulted in major shifts in membership functions that were consistent with predictions.