1What are the two main problems with applying traditional cognitive architectures to interface design?
They require hand-coded production rules (making strategy selection unclear) and are brittle (rules must be rewritten for changes in design/context).
2What are the two general sources of bounds in computational rationality?
Internal bounds (cognitionody limitations) and External bounds (physical world/device design).
3What are examples of internal bounds in the driving/scooter case study?
Limited peripheral vision, uncertainty about lane position, and noisy motor control.
4What are examples of external bounds in the driving/scooter case study?
Phone holder position, traffic density, and road curvature.
5How is policy derived in Computational Rationality compared to classical architectures?
In CR, it is derived via optimization (e.g., RL) over a POMDP with bounds; in classical architectures, it is hand-coded by the researcher.
6What is the role of Cognitive Science in the computational rationality framework?
It identifies processing limits and adaptation phenomena to justify the bounds used in models.
7What is the role of Machine Learning in the computational rationality framework?
It provides the formal framework (POMDP), bounded optimality notions, and algorithms (RL) to compute policies.
8Why are fixed-strategy models insufficient for interface design?
They cannot capture how behavior changes with design/context, failing to distinguish between designs or explain individual differences.
9What are the four general areas requiring adaptation in HCI (Section 2)?
Memory (chunking), Perception (eye movements), Motor (speed-accuracy trade-offs), and Environment/Ecology (traffic/workload).
10Why must theories explain "why" strategies emerge rather than just "how"?
To link behavior to reward maximization under bounds, enabling prediction of adaptation to new designs.
11In the Computational Rationality framework (Figure 2), what does the "Internal Environment" encode?
Mental states (percepts, memory, emotion) and their dynamics, including individual capacity limits.
12How does the Computational Rationality POMDP differ from a classical POMDP?
CR POMDPs explicitly model human cognitive bounds (internal environment) and define rewards over internal states, whereas classical POMDPs focus on external environments.
13What are the five theoretical commitments of Computational Rationality?
1. Agent solves bounded optimality problems. 2. Internal env encodes mental statesounds. 3. External env includes device/ecology. 4. Interaction is a source of bounds. 5. Subjective reward function over internal states.
14What are the elements of a POMDP tuple?
States (S), Actions (A), Transition function (T), Reward function (R), Observations (Ω), and Observation function (O).
15What are the four classes of bounds defined in Computational Rationality?
Time (duration), Noise (sensory/motor variability), Uncertainty (imperfect knowledge), and Capacity (limits on resources).
16What is the core aim of Cognitive Computational Neuroscience (Kriegeskorte & Douglas)?
To build task-performing computational models that are both cognitively meaningful and neurobiologically plausible.
17What is the critique of "20 questions with nature" in the context of CCN?
Testing isolated verbal hypotheses or mapping regions without building task-performing models fails to reveal mechanistic explanations of cognition.
18What are the three analysis families for "from experiment toward theory" in CCN?
Connectivity models, Decoding models, and Representational models.
19How do neural network models bridge theory and experiment in CCN?
They perform complex tasks with internal representations that can be compared to brain activity (e.g., ventral stream) and behavior.
20How does Cognitive Computational Neuroscience view Marr's levels?
They cannot be cleanly separated; understanding cognition requires integrating computational, algorithmic, and implementation levels.
21What is "adversarial cooperation" in the context of CCN?
A practice where different labs collaborate to challenge and falsify models using shared tasks and datasets.