Carl Westin

72 Source bias effects Information about respective source was intentionally minimized and limited in or- der to retain strict experimental control, to avoid confounding information that could influence participants’ attitudes. • Human source: All resolution advisories suggested in this session are made by an air traffic controller. • Automation source: All resolution advisories suggested in this session are generated by automation. In addition, a reminder of the ‘active’ source was stated in the advisory agree- ment dialog window that appeared together with an advisory and enabled partici- pants to either ‘accept’ or ’reject’ the advisory. The two dialog windows are shown in Figure 4-3. The advisory source was labeled either Automation or Human , with the latter referring to an air traffic controller. To further separate participants’ per- ception of the two windows, they were shaded differently with the ‘Automation’ window in dark grey and ‘Human’ window in light grey. In addition, the dialog window contained the agreement rating scale and a countdown timer indicating time remaining until advisory expiry. Debriefings indicated that the source manipulation was successful and that par- ticipants believed they were receiving advice from either an automated system or another controller. Order of advisory source was varied between participants, with three participants receiving first the human source, then the automated source. Source: Automation Advisory agreement Neutral Agree Disagree Accept Reject 30 (a) Automation Source: Human Advisory agreement Neutral Agree Disagree Accept Reject 30 (b) Human F IGURE 4-3: Advisory agreement dialog window shown during automation source condition (a) and human source condition (b).