Unit 5: Introduction to Intelligent Agents - Practice Quiz

1 What is an intelligent agent in the context of software bots?

2 Through what mechanism does an intelligent agent typically interact with its environment?

3 What is the primary role of an 'assistive' AI agent?

4 Augmented agents are primarily designed to:

5 Which characteristic describes an agent's ability to operate and make decisions without direct human intervention?

6 When an agent responds in a timely manner to changes occurring in its environment, it is demonstrating:

7 An agent that continuously takes actions calculated to move it closer to a specific desired outcome is exhibiting:

8 In an AI Agent node architecture, what typically acts as the 'brain' that processes inputs and decides on the next action?

9 What is the first step in an agent's perception-action loop?

10 In the perception-action loop, after an agent processes perceived information and decides what to do, what is the final step?

11 In the context of AI agents, what does 'function calling' allow the agent to do?

12 Why do intelligent agents integrate with external tools (like calculators, web browsers, or databases)?

13 What is the primary purpose of maintaining 'conversation history' in an agent's memory?

14 Short-term context in an agent's memory system is primarily used to:

15 What is the primary function of a customer support agent?

16 A research agent is specifically optimized to:

17 What is a key difference between an autonomous agent and a traditional workflow?

18 Which of the following is a common agent workflow design pattern where a primary agent decides which specialized sub-agent should handle a specific task?

19 If an agent attempts to use a web search tool and the tool fails, what is a standard error management strategy?

20 'Graceful degradation' in agent error management refers to:

21 A software agent is deployed to monitor network traffic and flag unusual spikes. Which of the following best describes why this system is classified as an 'agent' rather than a standard script?

22 A coding agent suggests auto-completions as a developer types, but relies entirely on the developer to accept or reject the code. Meanwhile, another agent autonomously refactors entire codebases and submits pull requests. How are these two agents classified respectively?

23 A trading bot is programmed to maximize profit by the end of the day. Suddenly, the stock market crashes, and the bot immediately halts all purchases to prevent losses before recalculating its strategy. Which two agent characteristics are primarily demonstrated here?

24 Which of the following scenarios best illustrates the 'autonomy' characteristic of an intelligent agent?

25 In an AI Agent node architecture, what is the primary role of the 'Reasoning Engine' or 'Controller' node?

26 During a perception-action loop, an agent parses an email, decides it is spam, and moves it to the junk folder. If the 'perception' phase is parsing the email, what constitutes the 'action' phase in this specific loop?

27 Why is the concept of a 'loop' critical in agent design rather than a simple linear execution path?

28 When designing an LLM-based agent that utilizes function calling to check weather data, what exactly does the LLM generate to trigger the function?

29 An agent is equipped with a get_stock_price function. The user asks, 'What is the weather in Tokyo?' How should a properly configured function-calling agent respond?

30 When integrating an external calculator tool into an agent, what is the most crucial information the developer must provide to the agent to ensure accurate tool usage?

31 An agent needs to access a secure database via a tool. Where should the API keys and authentication logic reside?

32 Which of the following is a direct consequence of an agent relying solely on a rolling window of conversation history for its short-term context?

33 To manage context limits in a conversational agent, a developer implements a summarization technique. At what stage should this summarization typically occur?

34 A task-specific research agent is designed to compile a report on recent AI advancements. Which combination of tools is most vital for this specific agent?

35 Unlike a general-purpose conversational agent, a customer support agent is highly task-specific. Which design constraint is most appropriate for a customer support agent?

36 When designing an automated pipeline for processing payroll, a developer chooses a standard workflow over an autonomous agent. What is the most likely reason for this choice?

37 In the 'Evaluator-Optimizer' agent design pattern, how do the two components interact?

38 An enterprise employs a 'Router' design pattern for its AI system. A user asks for technical troubleshooting. What is the primary function of the routing agent?

39 An agent attempts to call an external API, but the API returns a '429 Too Many Requests' error. What is the most robust error management strategy the agent's architecture should employ?

40 To mitigate the risk of an agent repeatedly failing in a continuous loop when a tool is broken, which mechanism is best introduced into the perception-action loop?

41 In an algorithmic trading agent, a sudden market crash triggers a hard-coded stop-loss protocol, overriding the agent's long-term portfolio optimization objective. Which conflict in agent characteristics does this scenario most accurately illustrate, and how is it typically resolved in architecture?

42 When designing an AI Agent node architecture to process high-throughput, asynchronous events, which component combination best prevents blocking the central LLM orchestration node while maintaining state consistency?

43 An agent operating in a partially observable environment enters an infinite perception-action loop because its sensory input does not reflect the changes caused by its actions. Which of the following introduces the most robust architectural fix for this specific failure mode?

44 An LLM-based agent uses function calling to interact with a database. It repeatedly hallucinates required parameters that do not exist in the JSON Schema provided in the system prompt. What is the most effective programmatic intervention to correct this at runtime without retraining the model?

45 When integrating a stateful external tool (such as a headless browser) with a stateless LLM agent, a critical challenge is maintaining the session context across multiple tool invocations. Which design pattern effectively bridges this gap while minimizing the context window overhead?

46 In an agentic system utilizing a sliding window for short-term context, crucial initial instructions are being pushed out of the context window during long conversations. Which architectural pattern best resolves this while maintaining strict context size limits?

47 A financial institution needs a system to process loan applications. The process involves strict regulatory compliance checks, deterministic math calculations, and fixed approval routing. Why is a standard workflow engine vastly superior to an autonomous AI agent for this specific use case?

48 The ReAct (Reasoning and Acting) design pattern is highly effective but can suffer from cascading failures if a single tool returns an ambiguous error. How does the Plan-and-Solve (Plan-and-Execute) pattern structurally mitigate this specific vulnerability?

49 An agent encounters an API rate limit (HTTP 429) during a multi-step data extraction process. To prevent systemic failure and context bloat, what is the most optimal error management strategy implemented at the agent orchestration layer?

50 An augmented agent is deployed alongside a human software engineer. Unlike a purely assistive agent (e.g., a standard chat copilot), the augmented agent autonomously monitors background CI/CD pipelines, identifies build failures, and drafts pull requests to fix them, pausing only for human approval. What defines the critical boundary between assistive and augmented in this scenario?

51 A task-specific research agent is designed to synthesize academic papers. When confronted with two highly cited papers presenting contradictory empirical results, what mechanism ensures the agent avoids generating a confidently incorrect, one-sided summary (hallucination bias)?

52 In a conversational memory system, as the dialogue history grows, relying solely on exact token truncation leads to a loss of core entity context. Which strategy leverages vector databases to provide the LLM with relevant historical context without violating short-term token limits?

53 The concept of 'bounded rationality' in intelligent agents suggests that an agent's decision-making is limited by computational power, time, and available information. In the context of LLM agents, which of the following is a direct manifestation of bounded rationality?

54 Consider an LLM instructed to call a function transfer_funds(source, destination, amount). The user says, 'Send $50 to Alice.' The LLM triggers the function but leaves the source parameter null because the user didn't specify it. What is the most robust architectural approach to handle this missing slot?

55 In a graph-based AI Agent node architecture, what is the primary computational advantage of structuring the agent's reasoning process as a Directed Cyclic Graph (DCG) rather than a linear chain?

56 When integrating arbitrary code execution tools (e.g., Python REPL) into an agent, security is a paramount concern. Which of the following isolation strategies provides the strongest security guarantee while still allowing the agent to perform data analysis tasks?

57 In a multi-agent workflow utilizing the 'Supervisor' design pattern, what is the specific role of the Supervisor node when resolving a complex, multi-domain user query?

58 A customer support agent is handling a return request when the user suddenly asks, 'By the way, do you sell hiking boots?' and then says, 'Anyway, back to my return.' How should the agent's architecture handle this sudden intent switch without losing the state of the return process?

59 You are tasked with building a system to perform Exploratory Data Analysis (EDA) on newly acquired datasets where the schema, data types, and required statistical tests are unknown beforehand. Why is an autonomous agent architecture more suitable than a traditional workflow architecture for this scenario?

60 In a heavily delayed perception-action loop, an agent executes an action (e.g., launching an advertising campaign) but only receives environmental feedback (e.g., conversion metrics) hours later. Which mechanism is essential to prevent the agent from thrashing (rapidly executing contradictory actions) during the delay period?