Thirty-seven intelligence officers completed two replications of tasks designed to investigate their subjective, quantitative interpretations of the source reliability and information accuracy (plausibility) rating scales. In judging a report, subjects were influenced much more by the accuracy rating of the report's content than by the reliability rating of the report's source. The mean probabilities assigned to the truth likelihood of reports described a linear relationship between rating level and probability for each scale. Most subjects were unable to treat reliability and accuracy independently; for these subjects, the higher a report's reliability rating, the higher the accuracy rating expected, and vice versa. Subjects were relatively consistent in their interpretations, but marked differences between subjects were observed. Structural inadequacies of the scales are pointed out and the development of a single-dimensional, quantitative scale is recommended.

The effects of information conflict, the frequency with which the information is updated, and the reliability of the information sources upon information seeking and decision behavior were studied. Sixteen U. S. Army enlisted men performed a computer-controlled task in which they had to request updated enemy-position reports from three information sources to determine which of eight locations was the target of agradual enemy advance. Forty problems, presented twice each, were arranged in two within-subject factorial designs-with source reliability varied between and within problems, respectively. Subjects were permitted to make up to three decisions on each problem, with correct decisions rewarded and incorrect decisions penalized according to a non-linear payoff function. Results from analyses of variance indicated that less informution was sought prior to an initial decision (a) for higher than for lower reliability conditions, ( 6 ) as update frequency decreased, and (c) (IS degree of conflict increased. When all sources were of higher reliability, accuracy of decisions was higher; but, in general, accuracy was complexly affected by conflict and update frequency. These results were discussed in terms of the cost of requesting information and the probability that an update would contain accurate information. Information sampling and decision latencies as well us other indices of performance were also evaluated and discussed.

Nineteen subjects were given a multinomial Bayesian task in which they had to determine which of four populations was being sampled. Each sequentially drawn data item was described on one of three dimensions of differing diagnosticity; knowledge of the given dimension was presented and the subject could choose to purchase or pass it up. Sixty computer-generated problems were administered in a 2 times 3 within-subject factorial design. A problem contained either 16 or 24 total items of information, and subjects were given sufficient resources to purchase either 50%, 75%, or 100% of the available information. A second study incorporating the same task, involving 20 subjects, and employing a one-factor design was carried out to compare available information limits that were specified either exactly (12 and 24 items) or within a range (12 to 24 items). Results indicated that: (1) significantly more information was sought as both amount of available information and resources increased; (2) as response uncertainty was reduced by the acquisition of additional information, accuracy generally increased; (3) when more information was available, subjects were more efficient in selecting data items; and (4) there was no significant difference in information seeking or accuracy between the average of the two exact-limit conditions and the range-limit condition of amount of available information.

No abstract