Agent Interactions
Structure and Architecture
Automation in UI Design
AI Decision-Making and Social Behavior
Did you pay attention?
100

What national holiday was used to simulate emerging social behaviors?

Valentine's Day

100

What large language model was used in the agent architecture?

ChatGPT

100

This principle ensures that an automated system assists users only when necessary rather than overwhelming them.

What is selective automation?

100

Agents rely on this cognitive-like process to interpret observations and decide on actions.

What is perception-based reasoning?

100

Name one of the authors of the papers we read this week.

(1) Joon Sung Park, (2) Joseph C. O’Brien, (3) Carrie J. Cai, (4) Meredith Ringel Morris, (5) Percy Liang, (6) Michael S. Bernstein, (7) Eric Horvitz 


200

True or false: agents can change the states of their environments in the experiment.

True

200

Each agent is initially defined in the sandbox as a paragraph written in natural language called a __?

Seed Memory

200

A good mixed-initiative interface should minimize this type of unnecessary disruption to users.

What is cognitive overload?

200

When agents communicate with one another, they use this structured exchange of information.

What is agent dialogue?

200

What video game did Denish reference in his response to the Generative Agents paper?

Hay Day

300

Agents in the sandbox interact with their world through their actions, and interact with other agents through __?

Natural Language

300

Name one common error that the agent simulations exhibited in the experiment (and where agents can plan their days).

(1) Agents don't recall the correct memories, (2) add embellishments to their memories, (3) overly formal speech from the LLM

300

To determine whether to intervene, mixed-initiative systems often rely on this branch of mathematics.

What is probability theory?

300

In AI systems, this principle prevents an agent from taking actions that contradict its previous behavior or objectives.

What is consistency enforcement?

300

Michael Bernstein received his PhD from this Boston school

What is MIT

400

To evaluate believability, the researchers crowd-sourced human evaluators from what crowd-sourcing platform?

Prolific

400

When perceiving observations in their environments, what core concept prevents agents from consecutively repeating the same actions?

Planning

400

One benefit of mixed-initiative interfaces is reducing the effort required for users to perform tasks, a principle known as this.

What is efficiency enhancement?

400

By analyzing previous experiences, agents can refine their decision-making over time using this approach.

What is reinforcement learning?

400

Name the exact title of one of the papers we read this week.

(1) Generative Agents: Interactive Simulacra of Human Behavior

(2) Principles of Mixed-Initiative User Interfaces

500

What are the two ways a user can interact with an agent?

1. As an interviewer

2. As the agent's "inner voice"

500

What three main components were used to simulate believable behavior in the agent architecture?

1. Memory Stream

2. Reflection

3. Planning

500

Adaptive automation relies on continuously adjusting assistance levels based on this aspect of user behavior.

What is real-time feedback?

500

This technique allows agents to predict the outcomes of their actions before executing them

What is forward modeling?

500

What is Michael Bernstein Professor of?

What is Computer Science