The concept of Behavior Description Interviewing has been developing for over 80 years.
Alfred Binet, in 1919, used interviews to assess the intelligence of five children. The three teachers who interviewed the children were confident with their assessments; their assessments, however, were contradictory. H. L. Hollingsworth, in 1922, studied the rankings given in interviews to applicants applying for positions in the Army. Several officers interviewed the same applicants; once again, not many of the officers gave the applicants similar rankings. One applicant, for example, was given the top ranking by one officer and a ranking of 57 by another.
The need for more accuracy was clear. In attempts to increase interview validity (accuracy), researchers began to study the value of adding structure to an interview. In these studies, each candidate was asked the same questions. One study reported a validity co-efficiency of .68 for job performance ratings and .43 for the amount of time a person would stay with the company. A second revealed .61 for the length of time employees stayed. Another study showed a .71 correlation between the ratings given in structured interviews and supervisor on-the-job ratings.
The value of structure was confirmed in these studies as well as later ones. In addition, researchers began looking for other ways to increase interview validity.
One extensive study examined the relevancy of favorable and unfavorable characteristics of the applicant. Managers were asked to evaluate several hypothetical applicants, half of whom were given a complete job description while the other half were given no job description at all. Managers who were given complete job descriptions were less influenced by irrelevant details than those working without job descriptions.
In 1965, J. B. Mass asked interviewers familiar with the characteristics of top performers to list behavior-based examples of low, medium and high performers. Another set of interviewers was given these examples and sorted them in the same categories without knowing how the other interviewers had grouped them. These examples were then used in rating applicants responses. Two studies reported .58 and .69 correlations, respectively.
In two studies done in 1980 headed by Latham, Saari, Pursell and Campion, interviewers used questions developed from critical incidents that asked candidates how they would perform in a specific situation. Interviewers were also given instruction on how to rate the candidates answers. The reliability scores of these studies were .71 and .67, and the correlations with job performance measures were .46 and .60. This type of interviewing became known as situational interviewing.
Upon review of these studies, Tom Janz took elements of the situational interview and developed a new type of interviewing: Behavior Description Interviewing. Both situational interviewing and Behavior Description Interviewing (BDI) base their questions on critical incidents, but the difference lies in the type of questions asked. Situational interviewing focuses on what applicants would do in a specific situation, whereas BDI asks applicants to relate what they did do in a specific situation, that is, BDI focuses on actual behavior. Janz created the phrase, The best predictor of future performance is past performance in similar circumstances, which is now associated with behavior-based assessment throughout the world.
In 1982, Janz compared traditional, unstructured interviews to Behavior Description Interviews in a study of teaching assistants (TAs). One group of eight senior business students was trained in traditional interviewing techniques, and another group of eight students learned how to give Behavior Description Interviews. Fifteen TAs were interviewed and rated four times, twice using the traditional method and twice using BDI; the criterion for the study was the ratings the TAs received at the end of the semester by students. Forty-five of the 60 interviews were properly recorded and then analyzed. There was better agreement between those ratings produced in traditional interviews (.71 vs. .46), but the BDI ratings were significantly more accurate (.54 vs. .08) in predicting TA performance ratings.
Three years later, Christopher Orpen performed a similar study involving life insurance sales. Each of the interviewers were given the same amount of training and similar opportunities to practice their skills after being randomly assigned to either BDI training or traditional, unstructured interview training. The two criteria of the study were the dollar value of policies sold and supervisors ratings. Upon conclusion of the study, Orpen reported a validity for the supervisors ratings of .56 for BDI and .08 for traditional, unstructured interviews. The interview validity (the average of two interview ratings) of the sales value was .72 for BDI and .10 for traditional, unstructured interviews.
Studies continue to confirm the validity of assessing actual behavior through Behavior Description Interviewing. No other interviewing method has proven to be as accurate in determining performance effectiveness.