Cognitive Interaction design

What is distributed cognition, and which cognitive system does it investigate?

Distributed cognition studies how coordinated people and artifacts jointly perform a cognitive task. The cognitive system is the whole task-performing ensemble, not a single individual. (Course notes, pp. 26–27.)

How is the boundary of a distributed cognitive system determined?

The boundary is defined by the resources deployed in solving the cognitive task. Relevant people, artifacts and representations must be included even when they extend beyond one individual or device. (Course notes, pp. 27.)

Use ship navigation to explain how cognition can be distributed.

Crew members measure and communicate bearings, and representations on a chart help determine the ship’s position. The coordinated system performs a computation that cannot be understood by studying only one participant. (Course notes, pp. 26–27.)

What is assistive redundancy, and why can it help a distributed cognitive system?

Assistive redundancy means that information is represented or transmitted in multiple forms, locations or channels. These overlapping resources can support reliable performance and error detection and correction. (Course notes, pp. 27.)

Explain the central claim of Clark and Chalmers’ extended-mind hypothesis.

The extended-mind hypothesis claims that parts of the environment can constitute parts of an individual’s cognitive processes when they play the relevant functional role and are actively coupled to the agent. (Course notes, pp. 33.)

Which criteria for external beliefs does the course text emphasize?

The notes emphasize accessibility without difficulty, regular reliance when relevant, and automatic trust in the retrieved information. A footnote also discusses prior conscious endorsement. (Course notes, pp. 33, inklusive fotnot 18.)

How do distributed cognition and extended mind differ in their starting point?

Distributed cognition starts with the system and asks how its components jointly cognize. Extended mind starts with an individual and asks which integrated external resources can constitute parts of that individual’s cognition. (Course notes, pp. 33.)

Explain the functional principle behind extended mind using an external memory.

An external process can count as cognitive when it plays a role that we would recognize as cognitive if performed internally. Otto’s notebook illustrates external memory integrated through accessibility, reliance and trust. (Course notes, pp. 33.)

Define the physical stance and give a useful example.

The physical stance explains and predicts behavior from physical states and natural laws. For example, a conducting object touching two contacts can cause a short circuit. (Course notes, pp. 19.)

Define the design stance and give a useful example.

The design stance predicts behavior from the artifact’s intended function: it should do what it was designed to do. A toaster is expected to toast bread when operated as intended. (Course notes, pp. 19.)

What are the central assumptions of the intentional stance?

The intentional stance attributes beliefs, desires and rationality to a system and uses these assumptions to explain and predict its behavior. It is useful when this strategy successfully accounts for what the system does. (Course notes, pp. 19–20.)

Can the same artifact be understood using different stances? Explain.

Stance choice is pragmatic. Depending on the situation and explanatory purpose, we may shift between or combine physical, design and intentional explanations of the same system. (Course notes, pp. 20–21.)

Explain the conversation metaphor and give an example.

The conversation metaphor treats interaction as communicating with an intermediary that carries out the requested action. For example, the user asks an assistant to move a file into a folder. (Course notes, pp. 19.)

Explain acting-in-a-world and give an example.

In acting-in-a-world, the user directly perceives and acts on objects in a model world. Dragging a visible file into a folder is an example of performing the action oneself. (Course notes, pp. 19.)

State the central difference between conversation and acting-in-a-world.

Conversation involves communicating with an intermediary, whereas acting-in-a-world involves the user directly perceiving and affecting the represented world. The contrast concerns the mode of interaction. (Course notes, pp. 19.)

Why can conversational interfaces create inappropriate expectations?

Conversational behavior may invite the intentional stance and expectations of coherent understanding. If the system is only locally or narrowly capable, users may predict its behavior incorrectly. (Course notes, pp. 21.)

What is internal complexity, and how does it relate to i-states?

Internal complexity concerns the artifact’s internal workings and its range of functionally important internal states, or i-states. (Course notes, pp. 36.)

Distinguish external complexity from interaction complexity.

External complexity concerns the artifact’s interface with its environment and its operationally important e-states. Interaction complexity concerns the dynamic relations between what the environment does and what the artifact does. (Course notes, pp. 36–37.)

What is mediated complexity? Give a concrete example.

Mediated complexity is located in the environment rather than in the artifact’s interior or exterior, yet affects interaction. People, furniture and routines in a shared corridor can contribute to it. (Course notes, pp. 37.)

Explain how removing buttons can redistribute complexity.

Removing buttons can reduce external complexity while increasing interaction complexity through more complicated sequences or combinations of actions. Complexity is redistributed rather than necessarily removed. (Course notes, pp. 37.)

Are the four loci sufficient for a full understanding of interaction complexity and quality? Motivate a position.

The loci are useful for locating complexity and describing trade-offs. They are insufficient as a complete account of difficulty or interaction quality, because users, training, goals and context also matter. This is an argued interpretation, not an official marking answer. (Course notes, pp. 37; fördjupning 44–45.)

What is automation, and what shifts in complexity may accompany it?

Automation transfers some execution or control to the artifact. User-facing external and interaction complexity may decrease while internal complexity increases, possibly together with complexity at the interface to the rest of the world. (Course notes, pp. 37–38.)

Discuss a cognitive benefit and a cognitive cost of automation.

Automation can reduce low-level cognitive effort and free attention for higher-level control. Potential costs include loss of direct control, deskilling and reduced contribution of context-sensitive human judgment. (Course notes, pp. 38.)

Name the four motives for automation listed in the notes.

The four motives are to disempower, relieve, outperform and augment the user. They can overlap and involve different distributions of control and capability. (Course notes, pp. 38.)

Why is cognitive interaction design particularly important for social robots?

Everyday robots share dynamic environments with people and may act autonomously. Cognitive interaction design supports intelligible behavior, appropriate expectations, communication and coordination between agents. (Course notes, pp. 61–64.)

Define i-states, e-states and w-states, with one robot example for each.

An i-state is a functionally relevant internal artifact state, such as a stored goal. An e-state is an external state of the artifact, such as a status display. A w-state is a relevant world state, such as a person’s position. (Course notes, pp. 35–36.)

How would you design requirements for a corridor interaction using i-, e- and w-states?

Specify the relevant world situations and success criteria. Require sensing to update i-states, relevant e-states to communicate behavior, and actions to respond to changed w-states. Include recovery when internal beliefs disagree with the world. (Course notes, pp. 35–36 och 61.)

What is common ground, and why does it matter in human–robot interaction?

Common ground concerns shared understandings, beliefs and values, including information about the extent of that sharing. Establishing it can make cooperation and communication more efficient and less vulnerable to misunderstanding. (Course notes, pp. 63–64.)

What does it mean to view interaction from the cognitive perspective?

The cognitive perspective considers an agent in an environment and criteria for its success. It asks how information acquisition, representation, processing and action allow the situated agent to act successfully. (Course notes, pp. 2–3.)

Does cognition mean only conscious thought? Explain.

Cognition is broader than conscious thought. It includes perception, memory, problem solving, decision-making and action, across conscious and non-conscious processes. Perception and action depend on the agent’s situation. (Course notes, pp. 2–3.)

Explain design space using dimensions, points, constraints and trade-offs.

A design space represents possible designs along dimensions such as size, cost and battery life. A particular design is a point, and constraints and dependencies restrict the feasible region and reveal trade-offs. (Course notes, pp. 52.)

Give a design-space trade-off and explain why mapping the space is useful.

The example shows that design dimensions can depend on one another. Mapping feasible combinations helps compare alternatives, understand trade-offs and identify promising unexplored designs. (Course notes, pp. 52.)