Abstract
Keywords
It is widely recognised that applying Human Factors (HF) knowledge to the design of systems provides important benefits both for users and for the overall cost effectiveness of acquisition. A common focus of HF design effort is the Human-Machine Interface (HMI) of a system, that is, the way information is presented and how a user can interact with the system.
When users learn to use a new system, there is a demand of effort, cost and time associated with training how to use that specific HMI. For example, learning what types of HMI widgets one can interact with, for example, menus, dialogues, and where relevant commands can be found. The design of the HMI may also influence how much refresher training is required to maintain HMI skill levels over time, for example, after periods of non-use.
The design and delivery of training within the military requires the organisation of scarce, high-demand resources such as training facilities and personnel. Any reduction in training required for a system, facilitated through HMI design, should translate to through life cost-savings, personnel and facility flexibility, and improved operational conduct.
This article discusses research undertaken to understand the scope for minimising the initial and ongoing training demand of military systems through improvements to HMI design. It initially explores the features of HMI which impact and support novice users training through design, how the technique of progressive disclose was investigated experimentally, as a potential method to support HMI learnability, and finally the researchers discuss and propose strategies and methods to support the way forward regarding HMI design to improve and maintain learnability.
BACKGROUND: REDUCING TRAINING DEMAND THROUGH DESIGN
Training demand refers to the learning required to operate a system through the HMI to a desired standard of performance; that is, to become and remain competent. Put simply, training demand is described as a function of: • What users must learn – the collection of knowledge, skills, and attitudes required for a given activity; • How long learning takes to achieve competence; and • How often re-learning must occur (Arthur et al., 1998).
Training represents a significant cost of ownership of any system, and many approaches to reducing the quantity or cost of training have been developed, some of which are based on the design and deployment of technology. For instance, simulation and display technologies such as Virtual Reality (VR) can provide a rich training environment.
However, of greater interest here are approaches to reducing training requirements by creating interactive systems and HMIs that are more learnable via their design, so that they support users in gaining proficiency (Grossman et al., 2009; Poretski & Tang, 2022).
A number of design strategies have been presented for addressing training demand by improving learnability; three broad approaches emerge: 1. The application of principles of ‘good’ design can make systems easier to learn as a part of improving usability; 2. Novel interaction technologies can (potentially) yield interactions that are more ‘intuitive’ or ‘instinctive’ and therefore present a lower learning burden; 3. Using design patterns or adding interface features that make the human-machine interface easier to learn.
Design Guidance and Principles to Enhance Learnability
The interaction design literature provides a rich source of guidance, heuristics, principles and patterns that capture good practice for creating good, usable interactions. Much of this guidance promotes design principles that improve usability by making systems that are easier to learn. For instance, the need for consistency appears in design principles by Norman (2013), Nielsen & Molich (1990), Wong (2020) and Shneiderman et al. (2016). By making a design consistent, the demands on the user to learn how to use the system, and therefore, the training requirement are reduced.
Subsequent authors have further explored the concept of consistency. For instance, Jimenez et al. (2011) identify different kinds of consistency: internal (or within-device) consistency refers to consistency with other applications developed for the same device; external (or across-device) consistency refers to consistency with other devices; and ‘metaphoric consistency’ refers to the use of metaphors or analogies to objects, attributes or relations in the world outside of computer interfaces.
Design Patterns and Strategies to Aid Learnability
The discoverability of a user interface (UI) is the ease with which users can find new features or functions of a system and learn to use the things that they find (Interaction Design Foundation, 2023); indeed, the term is sometimes used interchangeably with ‘learnability’. Good discoverability allows users to locate something they need, and thus enhances their ability to learn a specific task. Visibility of HMI design features, use of design affordances, user feedback, structuring the HMI design to match with user’s mental models and design consistency are some of the design principles that contribute towards discoverability of an HMI design. Importantly, discoverability and learnability complement each other. A similar approach is to design for the staged introduction to the user of interface features as a form of scaffolding to support learning.
Progressive disclosure is an interaction design pattern that sequences information and views or phases of interaction (Nielsen, 2006). Access to advanced or rarely used features is deferred to a later phase of interaction, making applications easier to learn. By limiting information and HMI design features presented at one time, the user choices and decisions are reduced. Progressive disclosure makes it easier for users to manage the complex HMI designs and makes them less error-prone, as they gradually learn to use the HMI through discoverability.
Multi-layered interfaces is one approach for achieving progressive disclosure, and supporting learnability, as proposed by Leung et al. (2010). Multi-layered interfaces can be designed to support learning such that novices first learn to perform basic tasks by working in a reduced-functionality, simplified layer (version) of the interface. By hiding the advanced settings to other layers, progressive disclosure helps users avoid errors and saves them the time they would have spent contemplating features that they don’t need. Once users have mastered this layer or require more advanced functionality, they can transition to increasingly complex layers and learn more advanced tasks. Similarly, Spannagel et al. (2008) employ a ‘training wheels’ approach by allowing learners to initially access a restricted set of features. Progressive disclosure has the advantage that it may be used in isolation without necessarily requiring training or tutorials. The ‘permissible’ navigation routes for any given context would be immediately obvious via the unavailability of irrelevant, not applicable or destructive actions for that context.
Nielsen (2006) has proposed the following guidelines for progressive disclosure: • Getting the right split between initial and secondary features; • Making it obvious how users progress from the primary to the secondary disclosure levels; • Avoiding multiple ways to progress to secondary options; and • Consider multiple secondary displays, each of which is revealed by a different control on the initial display.
Considering these guidelines, it is important to decide how to divide the primary and secondary features of the different layers of the interface: to understand which options are used together and which are better thought of as separate levels. The more features are deferred to another layer, the simpler the initial HMI-interaction will be. However, if the task is divided into too many stages or levels, users may get bogged down in excessive navigation.
To get the ‘right split’, task analysis and field studies will help provide insights into what users do and their requirements. If an existing system is re-designed, frequency-of-use statistics can help to prioritise features.
Of particular interest are design HMI interventions that make it easier for users to discover the features and functions of an interface in an efficient and effective manner. Achieving discoverability by progressive disclosure of interface features in a multi-layer interface presents a practical route forward.
EXPERIMENTAL INVESTIGATION OF PROGRESSIVE DISCLOSURE
An experiment was designed and conducted to investigate the potential learning benefits of a progressive disclosure interface multi-layer technique, for military participants, to aid learning a complex UI.
The experimental aims were to: 1. Assess the multi-layer technique for assisting users in finding important interface features and performing tasks, when learning a new interface; 2. Assess the impact of a multi-layer interface on users’ assessments of usability; and 3. Assess any trade-off between assisting users in finding task specific interface features, and supporting their awareness of a wider set of interface features, in a multi-layer interface.
Experimental Setup
A complex Military style UI application, The ArduPilot Mission Planner,
1
was chosen for the experiment (Figure 1), as it is a complex, real-time safety critical application that has many of the features and functions of military HMIs, a novel approach to the domain this research was conducted to inform. Mission planning task – experimental set up.
Mission Planner was customised and adapted to have two conditions: An Experimental multi-layered UI where features not needed for the task were ‘greyed-out’ and a control condition where all of the features were available (Figure 2). A between-subjects design was used to avoid between-condition learning effects. HMI elements within the application, with example ‘Greying Out’ menu commands.
Experimental conditions were allocated to participants via information technology experience questionnaires, which included familiarity with mission planner software for Unmanned Air Vehicles.
A hierarchical task analysis was conducted on the Mission Planner software, enabling temporal task sequences to be created which were mapped to beginner/basic and advanced UI functions.
Forty-two entry level military participants were given four Missions to complete (MODREC protocol MODREC 2075/MODREC/21). Missions 1 and 2 were orientation Missions to familiarise participants with the tasks and UI. A Basic Mission was then performed using the same tasks, but in a different map location. Finally, an Advanced Mission was performed using a slightly different task set and different map location. In the Advanced Mission, Full Interface (Full-Int) features were available to both the Experimental and Control Groups.
The System Usability Scale and a semi-structured interview (Experimental group only) were used to record participant preferences, alongside researcher observations. Objective measures included an end of experiment recognition test; number of researcher interventions required (to help participants); click-streamed log files (elements clicked and time clicked) and videos of users’ screens. Task performance was measured by recording the time taken to complete missions; the time taken to find ‘Used-before’ 2 functions and also the time taken relating to ‘Transfer findability’ 3 functions.
Findings
Statistical significances were not obtained between the conditions (in task performance or awareness of interface features); however unforeseen individual differences between subject performances tended to ‘mask’ experimental conditions effects. Nevertheless, data trends (quicker learning, fewer errors and fewer functions to navigate through) indicated that multi-layered interfaces can quickly ‘teach’ a user how to navigate through a complex UI when first learning it. It also has the potential benefit of preventing errors, destructive actions and mitigating users getting ‘lost’ in a UI. In addition, close observation of user behaviour indicated that users who methodically scanned menus were more adept at finding a function later compared to those who stumbled upon it by accident.
As a result of the individual differences, it could be argued that the focus should be on individuals (experience/mental models/metaphors) and their own control over the UI method employed, rather than trying to implement a one-size fits all solution. Figure 3 illustrates the wide variation in individual performances across experimental and control conditions. However there was an obvious trend that the Experimental condition (Progressive Disclosure) resulted in more consistency of performance between subjects. Between subject variation in the Experimental and Control Conditions.
GUIDANCE AND IMPLICATIONS OF THE STUDY
Improving HCI/HMI Learnability Through Design
The literature review, conducted as part of the research and detailed in the background of this article, highlighted how ‘good’ Human-Computer Interaction (HCI) design can help increase the usability of interfaces, with numerous articles demonstrating the benefit of particular HCI design features and a wealth of guidance available to support HCI development and evaluation.
However, the reviewed literature didn’t identify a link between HCI design features and reduced training. A relationship was identified between learnability and progressive disclosure; however, this was not validated through the experimentation, most likely due to individual differences being the dominant factor influencing user training and experimental performance.
Categorisation of HMI Interventions.
Interface Design and User Needs
Several interface priorities (Table 1) may conflict or require progressive implementation, for example, a dedicated application may use progressive disclosure to help a user learn the interface, then progress to full functionality once users feel competent. The user should also be able to ‘progress’ to full functionality once they feel comfortable as to where functionality is located (and return to it if refresher training is required).
Tailoring interfaces to users’ training needs, could be supplemented by protecting the ‘system’ against destructive user actions, whilst allowing the user to easily correct a mistake without having to re-input data. For example, validation of user inputs, consistency with earlier inputs and alerting users to inconsistencies. In addition, pre-populating data from earlier entries or through logical deductions due to selections made will also help to guide the user and reduce potential errors.
In addition, it is also pertinent to recognise that reducing training demand requires a holistic approach within the organisation within which it is to be implemented. User Centred Design practices help ensure design principles and guidance are implemented such as common language, symbology and interaction techniques and methods, thus supporting natural, intuitive and ‘good’ interface design. When a user is competent with a Microsoft Windows interface, they can use many Microsoft applications with a basic level of competence quickly, due to common interface and interaction styles. Adopting this approach in defence, could reduce training needs due to skill transfer and create a standard baseline mental model for applications.
Commonality in HCI/HMI to Support Leanability
Increasingly gaming controllers are used to control and navigate complex interfaces due to the vast number of personnel who are already competent with their mapped control functions and commands. Within the gaming world, players are able to navigate three-dimensional complex environments, whilst undertaking parallel tasks such as swapping weapons, and collecting other bonus items with one simple controller. The intuitiveness of such controllers, and the partially trained generation that are used to using them are already being tapped into as a mechanism to reduce training and is something that it is felt could be exploited further.
A single log-in, skill monitoring system could also be used to potentially track competence and focus training depending on their interface interaction style (such as exploratory, ability to scan, recover from errors, perceived mental models). An intelligent interface could also monitor user behaviour and could provide prompts to reduce the occurrence of repetitive errors.
Looking further ahead, Artificial lntelligence (AI) facilitated context sensitive-help could also be provided to the user, with Tool-Tips and a graphic of progress through multi-step processes. Allowing users to view the reasoning behind advice and reject it if desired should also be considered. Aligned to support functions, voice interaction is becoming more prevalent as a ‘natural’ way for humans to communicate with systems, but may not be practical in certain acoustic situations. Future experimentation should employ within-subjects designs, to reduce individual differences potentially masking experimental effects, and also explore the effect of motivation on learnability of the user interfaces.
Summary
The research detailed in this article explored the concept of minimising training via HMI design. Design features of HMI which influence and potentially support training novice users were examined though the literature review, which led to the technique of progressive disclose being investigated experimentally, as a potential method to support HMI learnability. Although individual differences tended to mask experimental effects, it was clear that individuals previous HMI experience, visual exploration technique and perceived task priority (e.g. speed vs. accuracy) all influenced ability to find a specific function. The best technique to support function finding will depend on context and learnability priority (e.g. quick learning vs. deep learning). Progressive disclosure would offer a significant advantage for quick learning and circumstances where users are swapping between applications and may forget previous learned pathways to functions. Further to this, the implementation of generic interface design heuristics, guidelines and principles will support learnability and thus should reduce training demand. Looking towards the future, novel AI and adaptive interfaces should be explored to understand how they can provide benefit to HMI design by understanding the end user further and adapting interfaces and interaction techniques to improve efficiency and the development of competence.
MODREC Sidebar
MODREC is the Ministry of Defence Research Ethics Committee (MODREC). It is an organisation which ensures all MOD funded or sponsored research involving human participants meets the required ethical standards (both nationally and internationally).
ORCID iD
Bob Fields https://orcid.org/0000-0003-1117-1844
Footnotes
