Kirstie said:
Pretty much every validity test that's been done on it has indicated that it has no temporal validity, disrciminant validity and is context specific. Put simply, that means:
- the results aren't stable over time
- the different roles aren't sufficiently distinct from each other to be meaningful and there's a lot of overlap
- the results vary according to the context in which you collect the data.
For recruitment this means that first, the results you get from the data collected when you recruit a person will not be robust enough to be transferrable to a later point in time when they are working for you; second, the kind of judgement opportunities they offer you are based on shaky criteria and finally the manner in which someone responds in a recruitment situation might be completely different to the manner in which they respond in a live work situation.
In short because the results are essentially unstable using it to make judgements about what a person might be like when making a decision as to whether to employ them or not is at best not particularly good practice and at worst unethical.
I bow completely to your superior knowledge here Kirstie and would be interested in what would be a good interview technique (if one exists)
I wonder more and more if a lot of interviews are actively bad, rather than just no good.
I do a lot of interviewing and I believe the process, whichever we use, is flawed. There is a tendency for the interviewers to favour some candidates (often probably based on ethnicity/sex/attractiveness etc) and the interview is then a process of confirming these initial prejudices.
I don't believe in interviews, the more I interview people!
I don't have data but I doubt there is a correlation between who we appoint and above average performance in the job (I'm interviewing doctors so they all have similar qualifications and training). It's rare to interview one who comes across well but is incompetant - most have a minimal level of competency.
I'm looking for people who are going to perform in an above average way at work and I don't know if that's possible.
(Unless we use a SirAllen type "interview from hell" and invite people to do random tasks for 8 weeks while shouting at them after each task)