Capítulo I: Transformaciones en la Paternidad
2. Rol que juega la mujer en la familia
Paper documentation is the most basic form of extrinsic user support and most modem systems and application software comes with at least one,
ii not a set or user manuals ana pernaps tecnmcai system documents tor tne system administrator/installer. User manuals are now commonly divided into 'getting started', tutorial and reference sections covering the various aspects of system or application functionality and use. The division between
tutorial and reference materials reflects the how to do it / how it works
distinction often found in discussions of user's procedural vs conceptual knowledge (Carroll, Aaronson 1988). It has been found that new users perform better when presented with as small a manual as realistically possible, rather than a lengthy one containing detailed explanations, and that performance in realistic tasks is improved if such so-called minimal manuals encourage the user to infer some of the information about the system instead of explicitly stating it. This phenomenon is probably due to the enhancement of conceptual knowledge that inferential reasoning about the system brings (Black & Carroll, 1987).
Although the quality of written manuals has shown some improvement in the last ten years, their effectiveness remains minimal for a number of reasons. Firstly, the use of a manual is often seen by users as a nuisance rather than a facility since it takes them away from their primary goal; the task they wish to perform using the computer system (Carroll, & McKendree, 1987). If they are used at all, it will often be as a last resort, a necessary evil, when all direct attempts to achieve the desired goal have met with failure. Secondly, manuals are static and cannot adapt to the users' level of knowledge or the specific nature of their query; bad habits, inefficient usage and lack of knowledge about facilities are not addressed and users have to recognise the need for help and be able to formulate it in a way that fits the organisation of the manual (Erlandsen & Holm: 1987). Thirdly, whereas the tutorial type of exposition of some user manuals is predicated on the idea of 'active learning', this approach actually does not alter the basically passive role of the user in following a set of instructions. Too often users complete the relevant task w ithout gaining any understanding of what it is they have done. Lastly, user manuals are often misplaced, lost or anonymously borrowed and thus cannot always be relied
upon as a source ox support. 5.6 On-line S upport
On-line documentation has the advantage that it will not normally be removed from access by the user and can be called upon at any time without the user having to move from the computer. However, it has a number of disadvantages. In common with paper documentation, it is static in nature and cannot adapt to the users state of knowledge and formulation of the problem. The user must be able to translate the problem to suit the structure of the on-line system in order to make best use of it. Also, in some systems it is necessary to scroll through large volumes of text in order to find the desired information.
Where on-line documentation allows the user to specify a topic about which help is sought, the topic typically has to be specified in system terms rather than from the user’s perspective of the problem. For example, the UNIX® 'man' command must be supplied with a proper UNIX® command and so the user must know which command relates to the question in hand. So-called 'context sensitive' on-line docum entation presents only information pertinent to the current state or mode of the system, thus limiting the volume of material requiring sorting and menu driven on line help systems can provide a more structured access to information but, again, it is difficult, if not impossible, to make these systems adaptive to particular situations (Houghton, 1984).
In a more active vein, some systems provide prompts to the user if the required input is not forthcoming. For example, VAX/VMS will prompt for missing parameters in a command. A more sophisticated kind of prompting is implemented in Interlisp’s DWIM (Do What I Mean) where incorrect input is interpreted via a spelling corrector and the intended input
presented to the user for confirmation {op. cit.). In some cases DWIM will
auto correct without consulting the user but this can produce unexpected results (Lewis & Norman, 1987).
5.7 Intelligent Help System s
The shortcomings of the kinds of user support described above have prompted many attempts to provide users with the kind of help facilities which might be expected from a human expert, without the associated costs. Some of the perceived qualities of such a system are the abilities to answer natural language questions submitted in the user's own terms, diagnose incorrect or inefficient user procedures and extend the user's knowledge of the system by introducing untried facilities.
There is, of course, a great deal of overlap between automated help and coaching or tutorial systems. Both are intended to promote in the user facility over a domain of knowledge or skill and in both, the topic within the knowledge domain addressed may be determined by the user, whether implicitly or explicitly. The main differences are that the knowledge
domain of a help system will always be internal, in that the information
provided concerns the use of some application, as opposed to being
external, where the inform ation concerns matters unrelated to the application in use. Thus, a user of a word processing package will receive
help about that package and its use, whereas the coaching within an
automated tutorial system will concern not the coaching system itself but some other sphere of knowledge or skill. Also, in the help system context,
the interaction is always bona fide, whereas a tutorial system may take the
user through specific training exercises and 'simulated' or 'dummy'
interactions.
Intelligent user support can also be classified according to a number of
other criteria. Systems may be passive, requiring the user to recognise the
need for help and to request it, or a ctive, where the need for help is
recognised by the system. They may be static, in that the content of the
information that they provide does not change as a result of the interaction, or dynamic, wherein the help provided is modified as a result of the interaction being monitored. Modifications may address both the content of information being accessed and its presentation. For example, poor user performance might require the help system to access low level knowledge
bases and to present detailed information, whereas a user designated an expert might need only brief messages on advanced topics. A mapping of these help system characteristics to users’ information requirements is also possible (Lutze, 1987). In general, the attribution of intelligence is taken as the ability to address wider issues than the literal meaning of a user's command or action and involves the use of knowledge-based techniques (Jones & Virvou, 1991).
The knowledge bases required by an intelligent help system concern the application itself, the user and help strategies for what the the user needs to be told. As well as some representation of the objects and available actions within an application, the application or user knowledge bases may also contain information concerning possible goals and plans, which may encompass a number of commands or actions. These plans may be predetermined or generated in real time by a planning component. The user model will typically model transient and persistent user knowledge, user misconceptions and attributes such as whether the user tends towards serial or holistic learning, for example. The help strategy knowledge base is responsible for providing information which guides the help system's side of the dialogue. The decision as to what to say and when to say it is determined partly as a result of the contents of this structure.
The provision of intelligent help is not only beset with a large number of issues and problems but is further confounded by the fact that very few of these can be tackled in isolation from the others. Most investigations of intelligent help address only a few aspects in the overall scope of the possible number of questions and do so in a restricted interaction context and there is little in the way of a generalised theory of intelligent help or exhaustive empirical data (Carroll & McKendree, 1987). As noted above, it is generally acknowledged that an intelligent help system needs knowledge of the user in the form of a user model but it is far from clear what user characteristics are appropriate or necessary. Would it, for example, be useful to model users' irascibility levels in order to determine the level of interruption by the help system that would be tolerated? W hat the system
determines as helpful information may be received by the user as an infuriating interruption. A possible corollary is that all messages are ignored which might lead to a catastrophic error. This might be termed the 'crying w olf problem.
Models of user knowledge need to address the fact that people do not leam merely by adding facts to a factual database. Learning is an active process wherein hypotheses are tested and revisions made on the basis of exploration. Prior experience is also brought to bear on the interpreting of novel situations (Carroll & Mack, 1984). More generally, it may not be possible to determine whether normative or stereotypical user models are adequate, except in relation to a particular help system function. The various issues in question have a great tendency towards interdependence.
There are often interpretational problems in modelling user intentions, plans and goals, owing to non-contiguity, non-linearity and ambiguity in authentic plan execution (Finin, 1983). Users will often change, suspend or abandon plans in mid flow and the relationship between plans and sequences of action is often many to many. Even human observers cannot always correctly interpret user actions, with the enormously greater interpretational facilities they have over any conceivable automatic system. The problem is exacerbated in situations where the task space is large and open-ended.
The active!passive distinction is itself perhaps oversimplified. Once help information has been presented there may well be a need to test its effectiveness or to allow the user to clarify or extend certain points. Studies of human help dialogues have found that users often propose answers to their problems by way of a request for help (Aarronson & Carroll, 1986). If these studies are relevant to the HCI context then a mixed initiative dialogue structure is suggested. Similarly, the level of control exerted by the system over the interaction has to be considered. A help system may merely advise or it may block further interaction until the user conforms to some recommendation or possibly auto-correct user errors. W hich of these strategies is appropriate and in which situations has not been
determined. The timing of help provision is also a critical factor, given the
dynamic nature of human action (ibid.).
On the positive side, it is reasonably clear that help systems which respond immediately to user errors are most effective, since the users attention is focussed on the topic concerned in the help message and the
correction is closely attached to the decision point (ibid.). Users tend to
ignore information which does not seem to them relevant to their current concerns. Thus it is also better to present information in terms which relate it to the user's current task or action. Users leam to recognise dead end situations if they have been allowed to encounter them, conforming to the exploratory characterisation of learning. They also appear to leam more effectively by pursuing their own goals rather than a sequence of instructions, also suggesting that help information should address the user's current task (Carroll & Mack, 1984). Of course, this will tend to limit users’ mastery of the functionality of a system to that pertaining to their current goals and plans. The aim of expanding user expertise beyond the scope of users' current goals and plans is an important one but one which lies more within the bounds of tutoring than assistance.