Aug 1, 2008
One of the main tasks in my day-to-day work as a science writer is tracking down experts. The web makes this much easier than it ever was for journalists in decades since. There are times when a contact in a highly specialist area does not surface quickly but there are also times when I know for a fact that I’ve already been in touch with an expert in a particular area but for whatever reason cannot bring their name to mind. Google Desktop Search, with its ability to trawl my Thunderbird email archives for any given keyword is a boon in finally “remembering” the contact.
However, finding just a handful of contacts from web searches, email archives and the good-old-fashioned address book pales into insignificance when compared to the kind of industrial data mining companies and organisations require of their “knowledge workers”.
According to Sharman Lichtenstein of the School of Information Systems at Deakin University, in Burwood, Australia, and Sara Tedmori and Thomas Jackson of Loughborough University, Leicestershire, UK: “In today’s highly competitive globalised business environment, knowledge workers frequently lack sufficient expertise to perform their work effectively.” The same concern might be applied to those working in any organisation handling vast amounts of data. “Corporate trends such as regular restructures, retirement of the baby boomer generation and high employee mobility have contributed to the displacement and obfuscation of internal expertise,” the researchers explain.
The team explains how knowledge is increasingly distributed across firms and that when staff need to seek out additional expertise they often seek an internal expert to acquire the missing expertise. Indeed, previous studies have shown that employees prefer to ask other people for advice rather than searching documents or databases. Finding an expert quickly can boost company performance and as such locating experts has become a part of the formal Knowledge Management strategy of many organisations.
Such strategies do not necessarily help knowledge workers themselves lacking the search expertise and time required to find the right person for the job, however. So, Jackson developed an initial expertise locator system, later further developed with Tedmori, to address this issue in an automated way. The researchers discuss an automated key-phrase search system that can identify experts from the archives of the organisation’s email system.
Immediately on hearing such an intention, the civil liberties radar pings! There are sociological and ethical issues associated with such easy access and searchability of an email system, surely? More than that, an expert system for finding experts could become wide open to misuse – finding the wrong expert – and abuse – employees and employers unearthing the peculiar personal interests of colleagues for instance.
The first generation of systems designed to find experts used helpdesks as the formal sources of knowledge, and comprised simply of knowledge directories and expert databases. Microsoft’s SPUD project, Hewlett-Packard’s CONNEX KM system, and the SAGE expert finder are key examples of this genre, the researchers point out. Such systems are akin to Yellow Pages and are essentially electronic directories of experts that must be maintained on a continual basis. They allow anyone with access to tap into expertise, but unless the experts keep their profiles up to date, they can quickly lose relevancy and accuracy.
Overall, when large numbers of employees are registered and profiles are inaccurate, credibility is rapidly lost in such systems which are increasingly ignored by knowledge seekers.
Second generation expertise locators were based on organisations offering their staff a personal web space within which they could advertise their expertise internally or externally. Convenient for those searching but again relying on the experts in question to keep their web pages up to date. Moreover, simple keyword matching when searching for an expert would not necessarily find the best expert because the search results would depend on how well the expert had set up their web pages and whether and how well they had included keywords in those pages. In addition, keyword searching can produce lots of hits that must then be scanned manually, which takes time.
The third generation of expert searching relies on secondary sources, such as tracking the browsing patterns and activities of employees to identify individual experts. Such an approach raises massive privacy concerns, even for companies with a strict web access policy. Activity on forums, bulletin boards, and social networks falls into this third generation approach.
The fourth generation approach mashes the first three and perhaps adds natural language searching again with various efficiency and privacy concerns. Again, it does not necessarily find the best expert, but often just the person whose data, profile, and web pages are optimised (deliberately or by chance) to reach the top slot in the search results.
An approach based on key-phrase identification in e-mail messages could, however, address all requirements but throws up a new wave of privacy concerns, which Lichtenstein and colleagues discuss.
There are several features of email that make it popular and valuable for organisational knowledge work, and relevant to to finding an expert:
- It attracts worker attention
- It is integrated with everyday work
- It provides a context for sense-making about ideas, projects and other types of business knowledge
- It enables the referencing of work objects (such as digital documents), and provides a history via quoted messages
- It has high levels of personalised messages which are appealing, meaningful and easily understood
- It encourages commitment and accountability by automatically documenting exchanges
- It can be archived, so providing valuable individual, collective and organisational memories that may be mined
- It facilitates the resolution of multiple conflicting perspectives which can stimulate an idea for a new or improved process, product or service.
All these factors mean that email could become a very useful tool for finding experts. Already many people use their personal email archives to seek out knowledge and experts, but widen that to the organisational level and the possibilities become enormous.
The researchers have developed an Email Knowledge Extraction (EKE) system that utilises a Natural Language ToolKit (NLTK) employed to build a key-phrase extraction “engine”. The system is applied in two stages, the first of which “teaches” the system how to tag the speech parts of an email, so that headers and other extraneous information become non-searched “stop words” within the email repository. The second stage extracts key-phrases from the searchable sections of an email once it is sent. This extraction process is transparent to the sender and takes just milliseconds to operate on each email. A final stage involves the sender being asked to rank each identified key-phrase to indicate their level of expertise in that key-phrase area. A database of experts and their areas of expertise is gradually developed by this approach. Later, employees searching for experts can simply consult this database.
The EKE system has been implemented at Loughborough University and at AstraZeneca in trials and found to be able to capture employee knowledge of their own expertise and to allow knowledge workers to correctly identify suitable experts given specific requirements. The researchers, however, highlights the social and ethical issues that arise with the use of such as system:
- Employee justice and rights and how these might conflict with employer rights.
- Privacy and monitoring, as there is more than a small element of “Big Brother” inherent in such a system
- Motivational issues for sharing knowledge, as not all those with expertise may wish to be data mined in this way, having enough work of their own to fill their 9-to-5 for instance
- Relationships, as not everyone will be able to work well together regardless of expertise
- Ethical implications of expert or non-expert classification, as the system could ultimately flag as experts those employees with little or no expertise.
- Deliberate misclassification of experts, as all systems are open to abuse and malpractice.
- Expert database disclosure, as such a comprehensive database if accessed illicitly by an organisation’s rivals could wreak havoc in terms of stealing competitive advantage, headhunting or other related activities.
Lichtenstein, S., Tedmori, S., Jackson, T. (2008). Socio-ethical issues for expertise location from electronic mail. International Journal of Knowledge and Learning, 4(1), 58. DOI: 10.1504/IJKL.2008.019737