Human Computer Interaction
ERCIM News No.46, July 2001 [contents]

Expert Contextual Online Help Strategies

by Noëlle Carbonell and Antonio Capobianco

Online help to the use of standard application software intended for the general public is still highly unsatisfactory. Despite the continuous efforts of researchers and designers over the past twenty years, most novice users still prefer consulting experienced users to browsing available online help or paper user guides. Specific training sessions and tutorials are still very popular among novices nowadays. A promising approach, based on the analysis of human expertise, is presented (project supported by the French Ministry of Defence and the CNRS). 

To be efficient, usable and actually used, online help has to overcome one major obstacle at least, that is the ‘motivational paradox’. Carroll and Rosson observed in the eighties that users in the general public are reluctant to explore new software and to learn how to use its functionalities efficiently. These users seem to be mainly concerned with achieving the tasks motivating their interaction with the software. Therefore, they are liable to ignore the facilities for autonomous learning provided by intuitive user interfaces in order to familiarise users with the operation of new software.

This paradox may also explain the failure of approaches which view online help as an interactive learning situation. In particular, it may account for the tendency of the general public to ignore online tutorials as well as online manuals in the form of databases, hypertexts or hypermedia.

To take this paradox into account, designers have to consider online help as a specific human-computer interaction situation. From this angle, contextual online help appears as an appropriate design framework. Providing users with the specific information they need to carry through their current task, and delivering this information at the right moment, contribute significantly to supporting users’ interactive activity, thus meeting their expectations.

However, selecting relevant meaningful contextual information and exploiting it appropriately are still crucial research issues, as illustrated by available implementations: products on the market are too crude to prove useful, and research prototypes too complex to be reliable, hence usable.

In order to improve the utility and usability of online help, we resorted to eliciting the help strategies of human experts from the analysis of expert-novice dialogues. Our study addresses two main issues:

Empirical Data and Analysis Method
We analysed expert-novice dialogues collected during an earlier study. Eight novice users performed twenty predefined formatting tasks on a given text, using Microsoft Word. They could communicate freely (over an intercom) with an expert user who helped them to carry out the prescribed tasks. Expert and novice were in different rooms; in one condition (A), the expert could view the novice’s screen (via Timbuktu), while in the other one (B) she could not.

Dialogues were tape-recorded, and the subjects’ screen displays videotaped. We used time-stamped written transcripts of the dialogues annotated with descriptions of the novices’ actions and their results on the user interface.

Our analyses are based on the labelling of all speech acts in the textual corpus, using specific taxonomies which were evolved from a preliminary survey of the dialogues.

Results and Interpretations
Requests for contextual information represent over 14% of the expert’s speech acts in condition B (vs less than 2% in condition A), and 87% of these requests aim at clarifying the software current state as displayed on the screen (54%), the progress of the current task execution (33%) or the subject’s current intention (10%). Therefore, any help system which aims at emulating human experts’ strategies should involve contextual information among its major knowledge sources.

A close analysis of the expert’s speech acts shows that she most often uses contextual information for selecting the help information she gives to novices (over 90% of her help speech acts in both conditions). The main source of dynamic contextual information is the progress of the current task execution (over 40% of her help speech acts).

These results indicate that the expert encourages novices to adopt an autonomous ‘learning by doing’ strategy for familiarising themselves with the operation of a new software, by helping them to achieve the tasks which motivate their use of the software.

Such a strategy, which differs from standard didactic strategies implemented in computer aided instruction, will be well accepted by users in the general public (cf. the motivational paradox). It will also be easy to implement, as it exploits a model of the novice’s current activity instead of a cognitive user model which is more complex to build, update and use.

Future Work
We are currently developing a prototype which will be used to evaluate (experimental evaluation with potential users) the usability and efficiency (in terms of learning) of this strategy, compared to other contextual help strategies.


Please contact:
Noëlle Carbonell and Antonio Capobianco — LORIA
Tel: +33 3 83 59 20 32, +33 3 83 59 30 83
E-mail: {Noelle.Carbonell,Antonio.Capobianco}