This workshop was given at both the South African Occupational Psychology Conference in Johannesburg (SIOPSA) and at the New Zealand Psychological Society Conference at Massey University, Palmerston North

Person-Target Profiling in I/O Applications: The Logic, Methodologies, and Optimal Contexts.
There is one key set of functions for Human Resource practitioners and I/O consultants alike: variously providing services, procedures, protocols, and advice on matters of personnel recruitment, selection for training programs, selection for promotion, and selection for entry into succession-planning programs. There are two kinds of approach to these issues. Both have at their heart the desire to select one or more individuals who will be most likely to succeed in whatever role or function for which they are being assessed and subsequently selected. The first and most prevalent approach is a largely subjective one. That is, processes, procedures, and selection strategies are constructed and administered via a mixture of expert opinion and experience, committee discussions, job-analysis, external consultant opinion recommendations, and the interpretation of whatever evidence exists to suggest that one set of procedures is likely to possess greater validity than another. The second approach attempts to reduce the sometimes excessive subjectivity in the first approach to a minimum, and attain the status of a “near-objective” approach to the selection problem. This approach treats the problem as one of simply finding the computationally defined “best matches” to predefined “targets” … in a word “Profiling”. However, three serious problems are encountered with this approach:

1. What kinds of information will be used to profile an individual?
2. What exactly will be constitutive of a “target” such that a match can be computed?
3. How exactly is a “match” to be computed?

 Before examining these three questions in detail, the logic and justifications for using profiling are initially presented and discussed. This is especially important when considering how psychometric test scores are used as part of any selection process, and how I/O consultants, companies, and computer-based administration test publishing companies have proposed various “profiler” solutions to clients. Using the Psytech GeneSys^TM Profiler system as an exemplar, this commercial state-of-the-art solution is examined both for its validity and limitations. When the logic and justifications for profiling are carefully compared to the actual profiling applications on the market, it is clear that there is a conceptual gap between the expectations and statements promoting the use of profiling systems and the current delivery of profiling solutions to the market. Identifying the existence of this gap simultaneously exposes the reasons for the gap.

However, whilst intellectually interesting and perhaps of some mild utility, such an analysis is of little practical value. This is where the workshop gets to the heart of the matter and builds the case for Smart Profiling, advocated by myself back in 1999. Smart Profiling might be considered by some to be a particular exposition of a class of methods subsumed under the nomenclature “data mining”; perhaps so. It is at this stage in the workshop where I answer questions #1 through #3, in detail. These answers variously utilise several computer programs and algorithms, both as didactic components of the workshop, as well as permitting real-time demonstrations of the implications and outcomes of key arguments, techniques, and methodologies being propounded in the Workshop. A little of what I discuss is already a commercial reality within StaffCV.com, a web-based job-board/recruitment process application. One of the key strategic issues to be discussed is how the HR-I/O consultants might proceed to utilise this technology within any medium size company or corporate. This is paradoxically both a barrier to its usage as well as the only means by which such usage can be made a reality! 

The aims of this workshop are to:

• provide the audience with a thorough exposure to the key issues of person-target profiling.

• present the arguments and empirical evidence in favour of the substantive more objective use of psychometric and other sources of data for optimal selection strategies.

• explore the somewhat limited approach that has been adopted within current approaches that utilise psychometric test scores for profiling.

•  provide all the essential equations and algorithms that are currently used in profiling systems.

• introduce a new class of matching indices known as “shaped-distance functions” and constrained-order distance functions.

• demonstrate how prone some coefficients are to generating nonsense “matches”.

• examine closely the problems with using homogenous-group mean scores for conventional target profile construction.

• provide real-world, working demonstrations, of Smart Profiling in action.

The materials disseminated to each member of the audience via CD will be a substantial pdf manual of profiling methods, equations, examples, and definitions. This also serves as the on-line HTML help to the computer program (also being distributed) used in the workshop that analyzes coefficient behaviour and expected distributions of 9 methods of computing person-target matches. The various Powerpoint presentations, search algorithm-schematics, and Delphi-7 computer-programs embodying such algorithms will also be distributed on this CD. Some time will be given to members of the audience who wish to install the software immediately on their laptops – and learn how to present their own datasets to these algorithms. However, it must be stressed that the Smart Profiling exemplar programs are just that. They are there to demonstrate what is possible, how it is achieved, the difficulties that one will encounter, and the kind of results one gets with this “information–rich” approach compared to current profiling methodologies. This is not a workshop designed to demonstrate or attempt to sell any particular commercial product. It is simply a detailed exposition of what happens when you combine new and established sets of methods and search algorithms with a broader information-base so as to produce more coherent matches of individuals to one or more sensibly constructed “targets”. The construction of those targets is also the focus of much empirical and pragmatic analysis in this workshop.

There are three workshop presentations - which may be downloaded here as pdf files.
Workshop 1, entitled: "Introduction" covers the basic concepts of 1, 2, and 3-D profiling, consists of 75 slides, and is 5.5Mb in size. It is this large because of all the graphics and equations contained therein.
Workshop 2, entitled "The Coefficient Distribution Analyzer" evaluates 9 profile matching coefficients in terms of their expected distributions of values under conditions of profiling artificial random and real personality data. It also evaluates the behaviour of the most promising subset of these using systematic degredation of a profile, and mapping the values of each given this step-by-step degredation. It consists of 72 slides, and is 862kb in size.
Workshop 3, entitled "Bigger Issues" discusses some of the strategic issues thrown up in presentations 1 and 2, especially regarding the tradeoff between increasing sophistication and actual utility. It consists of 20 slides, and is 325kb in size.
Workshop CD Layout one-pager pdf format document (74k) ... contains a listing of the files and subdirectories on the CD - with a few brief notes.

*Note, these presentations make use of programs (and avi files, documents, and papers etc.) which were run "live" during the workshop - and were given away on the workshop CD. Due to the size of the files (almost 200Mb), it is not feasible to distribute them on the web-page.