System Design

Human-Computer Interaction (HCI): Contemporary Approaches

As technology has advanced, the ways in which humans interact with computers have expanded far beyond the traditional keyboard and mouse. Contemporary HCI approaches include:

Touch Interfaces

Touch screens allow users to interact directly with what is displayed by tapping, swiping, pinching, and dragging on the screen surface.

Applications: Smartphones, tablets, self-service kiosks, point-of-sale terminals, interactive whiteboards

Advantages:

• Intuitive and natural – users directly manipulate on-screen elements
• No need for a separate input device (mouse/keyboard)
• Accessible to a wide range of users, including young children

Disadvantages:

• Lack of tactile feedback (no physical sensation of pressing a key)
• Imprecise for detailed work (e.g. fine image editing)
• Screen can become obscured by the user’s hand
• Hygiene concerns on shared public touchscreens

Gesture-Based Interfaces

Gesture recognition systems detect and interpret human body movements – hand waves, finger pointing, body posture – as input commands.

Applications: Gaming (e.g. Microsoft Kinect), smart TVs, surgical operating theatres (touchless control to maintain sterility), sign language recognition

Advantages:

• Hands-free interaction is possible
• Natural and immersive for gaming and entertainment
• Useful in environments where touching a device is impractical or unsafe

Disadvantages:

• Can be imprecise and unreliable
• Physically tiring for extended use (“gorilla arm” fatigue)
• Limited vocabulary of gestures compared to keyboard input
• Requires specialised sensors (cameras, depth sensors)

Voice Interfaces

Voice recognition systems convert spoken language into commands or text. Virtual assistants such as Siri, Alexa, and Google Assistant are prominent examples.

Applications: Virtual assistants, dictation software, accessibility tools, smart home control, car infotainment systems

Advantages:

• Hands-free and eyes-free operation
• Fast for simple commands and dictation
• Highly beneficial for users with physical disabilities
• Natural interaction for many users

Disadvantages:

• Struggles with accents, dialects, and background noise
• Privacy concerns (always-listening devices)
• Ambiguity in natural language (discussed in detail below)
• Difficult for complex or precise instructions
• Socially awkward in public or shared spaces

Virtual Reality (VR) and Augmented Reality (AR)

VR immerses the user in a fully computer-generated environment using a headset and controllers. AR overlays digital information on the real world, typically through a smartphone camera or smart glasses.

VR applications: Training simulations (surgery, flight, military), gaming, virtual tourism, architectural walkthroughs

AR applications: Navigation overlays, maintenance instructions, educational tools, retail (virtual try-on), medical imaging during surgery

Advantages:

• Highly immersive and engaging
• Excellent for training in dangerous or expensive real-world scenarios
• AR provides contextual information in real time

Disadvantages:

• VR headsets can cause motion sickness and eye strain
• Expensive hardware requirements
• VR isolates the user from the real world
• AR can be distracting and potentially dangerous (e.g. while driving)

Brain-Computer Interfaces (BCI)

Brain-computer interfaces detect electrical signals from the brain and translate them into computer commands. This is still a largely experimental technology.

Applications: Assisting paralysed individuals to control prosthetics or computers, neuroscience research, potential future gaming and productivity applications

Advantages:

• Can provide communication for users who cannot speak or move
• No physical input device required at all
• Potential for very direct and fast interaction

Disadvantages:

• Currently very limited in accuracy and speed
• Invasive BCIs require surgical implantation
• Non-invasive BCIs (e.g. EEG headsets) are imprecise
• Ethical concerns about brain data privacy
• Very early stage of development for consumer use

---

Natural Language Interfaces: Potential and Limitations

Potential of Natural Language Interfaces

Natural language interfaces represent one of the most intuitive forms of HCI because they allow users to communicate with computers in the same way they communicate with other people.

Potential benefits:

• No training required – users already know how to speak or type in their native language
• Accessibility – benefits users who struggle with traditional interfaces, including those with physical disabilities or limited computer literacy
• Speed – spoken commands can be faster than navigating menus or typing formal commands
• Versatility – can handle a very wide range of requests without the user needing to learn specific commands
• Multilingual support – systems can potentially support many different languages

Current Limitations

Despite significant advances in natural language processing (NLP), current systems have important limitations:

• Ambiguity – natural language is inherently ambiguous (see detailed analysis below)
• Context understanding – systems struggle with implied meaning, sarcasm, irony, and cultural references
• Complex queries – multi-step or conditional requests are often misinterpreted
• Domain knowledge – systems may lack specialist vocabulary for technical fields
• Conversational memory – maintaining context over a long conversation remains challenging
• Error recovery – when the system misunderstands, correcting the error can be frustrating

---

Problems of Ambiguity with Spoken Input

Spoken natural language input introduces additional layers of ambiguity beyond those present in written text.

Homophones

Spoken Word	Possible Interpretations
“there” / “their” / “they’re”	A place / belonging to them / they are
“to” / “too” / “two”	Preposition / also / the number 2
“write” / “right”	To compose text / correct or a direction
“flower” / “flour”	A plant / a baking ingredient
“sea” / “see”	A body of water / to perceive visually
“knight” / “night”	A medieval warrior / the time after sunset

A voice recognition system hearing “I want to book a flight to/two/too Nice/nice” must use context to resolve multiple ambiguities simultaneously.

Accent and Dialect Variation

Different accents and dialects can cause the same word to sound very different, or can make different words sound the same.

Problems include:

• Regional pronunciation differences (e.g. “bath” with a long or short ‘a’)
• Different vowel sounds across dialects
• Systems trained primarily on one accent may perform poorly with others
• Non-native speakers may have pronunciation patterns unfamiliar to the system
• Speed and rhythm of speech vary significantly between speakers

Background Noise

Real-world environments introduce noise that interferes with speech recognition:

• Office noise – other people talking, phone calls, keyboard typing
• Outdoor noise – traffic, wind, crowds
• Domestic noise – television, music, appliances
• Echo and reverberation – in large rooms or spaces with hard surfaces
• Multiple speakers – the system must identify which voice to listen to (the “cocktail party problem”)

Context Dependence

The meaning of spoken sentences often depends on context that the computer may not have:

• Conversational context – “Book that one” requires knowing what was previously discussed
• Situational context – “It’s cold in here” might be a request to adjust heating, or simply a statement
• Cultural context – idioms and colloquialisms vary between cultures and may not translate literally
• Emotional context – tone of voice, stress, and intonation carry meaning that is difficult for systems to interpret

---

Design Validation

Design validation involves several complementary checks:

Reviewing the Design Against the Specification

The design must be checked systematically against every requirement in the specification to ensure nothing has been missed or misinterpreted.

Methods include:

• Requirements traceability matrix – a table that maps each requirement to the corresponding design element, ensuring complete coverage
• Formal design reviews – meetings where the design team presents the design to stakeholders, who verify it meets their expectations
• Walkthroughs – step-by-step examination of the design, often using scenarios or use cases to check that the design handles them correctly
• Sign-off – stakeholders formally approve the design before implementation begins

Key questions:

• Does the design address every functional requirement?
• Does the design meet all non-functional requirements (performance, security, usability)?
• Are there any requirements that have been overlooked or misunderstood?
• Are there any design elements that do not map to any requirement (unnecessary complexity)?

Checking Appropriate Techniques Were Used

The design should use techniques and technologies that are appropriate for the type of system being built.

Considerations include:

• Has the appropriate development methodology been chosen (Waterfall, Agile, or a hybrid)?
• Is the chosen programming paradigm suitable (e.g. OOP for a system with many interacting entities)?
• Are the data structures and algorithms efficient for the expected data volumes?
• Is the database design normalised appropriately?
• Are appropriate security measures designed in from the start?
• Is the architecture scalable for future growth?

Checking User Interface Appropriateness

The user interface design must be appropriate for the intended users and their context of use.

Factors to consider:

Factor	Questions to Ask
Target audience	Who will use the system? What is their technical ability?
Accessibility	Does the interface meet accessibility standards (e.g. screen reader support, colour contrast)?
Platform	Is the interface appropriate for the target device (desktop, mobile, tablet, kiosk)?
Consistency	Does the interface follow established conventions and maintain internal consistency?
Error handling	Does the interface provide clear, helpful error messages?
Navigation	Can users find the features they need without difficulty?
Responsiveness	Does the interface adapt to different screen sizes and orientations?

---

Design Evaluation Criteria

Once a system has been built (or during design), it must be evaluated against a range of criteria to judge its quality and fitness for purpose.

Functionality

• Does the system perform all required functions?
• Are calculations and outputs correct?
• Does it handle all specified input types, including edge cases?
• Are all user roles and permissions implemented correctly?

Usability

Usability measures how easy and pleasant the system is to use.

• Can users learn to use the system quickly?
• Can experienced users complete tasks efficiently?
• Is the interface intuitive and consistent?
• Are error messages clear and helpful?
• Does the system meet accessibility requirements?
• Is user satisfaction high?

Performance

Performance measures how quickly and efficiently the system responds under various conditions.

• Does the system respond within acceptable time limits?
• Can it handle the expected number of concurrent users?
• Does it use system resources (CPU, memory, storage) efficiently?
• How does it perform under peak load conditions?
• Are response times consistent?

Reliability

Reliability measures how consistently the system operates without failure.

• How often does the system crash or produce errors?
• Can it recover gracefully from failures?
• Is data integrity maintained at all times?
• Does it include appropriate backup and recovery mechanisms?
• What is the mean time between failures (MTBF)?

Maintainability

Maintainability measures how easy it is to modify the system after deployment – fixing bugs, adding features, or adapting to changes.

• Is the code well-structured, commented, and documented?
• Are modules loosely coupled and highly cohesive?
• Can individual components be modified without affecting others?
• Is the system built using standard technologies and patterns?
• Is there comprehensive technical documentation?

Portability

Portability measures how easily the system can be transferred to a different environment (hardware, operating system, or platform).

• Can the system run on different operating systems?
• Is it dependent on specific hardware?
• Are external dependencies well-managed?
• Is the data stored in standard, transferable formats?
• How much effort is required to migrate to a new platform?

Cost-effectiveness

Cost-effectiveness evaluates whether the system provides good value relative to its cost.

• Does the system deliver the expected benefits?
• Are the development costs proportional to the value it provides?
• Are ongoing maintenance and operational costs reasonable?
• Does the system reduce costs or increase revenue as expected?
• How does the total cost of ownership compare with alternative solutions?