Quiz 1 Exam Prep - BD

26 vocabulary flashcards cover key AI-marketing concepts, risks, frameworks and technologies highlighted in the lecture notes to aid final-exam preparation.

Which marketing use of AI is least adopted despite high potential?

1. Personalized campaign creation 

Why: Only about 6% of firms reported using AI for advanced personalization like campaign creation, despite its high value

Which technologies combines multimodal inputs to assess customer emotion?

1. Neurodata Labs and Promobot’s emotion system 

Why: Neurodata Labs and Promobot use multimodal AI (voice, gestures, biometrics) to detect emotional states

What is “target leakage” in machine learning?

1. Including outcome-related data during model training

Why: Target leakage occurs when a model is trained on data that would not be available at prediction time, leading to inflated performance

Which of the following best illustrates the difference between statistical modeling and ML?

1. ML emphasizes predictive accuracy, statistical modeling emphasizes explanation

    

Why: Machine learning models are trained and applied with the objective of prediction, whereas most statistical models are focused on developing a description or explanation of the data

What is a key reason why AI and ML can outperform traditional statistical methods in marketing?

1. They do not impose rigid assumptions about the data 

Why: AI and ML are praised for their ability to learn from data without relying on the rigid assumptions that traditional statistics require

What is the primary reason AI has recently become more powerful for marketing applications?

1. Growth of Big Data, cheaper computing, and new AI techniques 

Why: The article attributes AI’s recent success to three converging forces: Big Data availability, cheap and scalable computing, and newer techniques like deep learning

1. What role does mechanical AI play in augmenting human marketers?

1. Performing undesirable, non-contextual tasks 

Why: Mechanical AI is used to take over repetitive, often undesirable tasks (e.g. cleaning, delivery)

What is a critical risk of using automated calibration without additional validation?

1. The model may overfit and be treated as a black box 

Why: Overreliance on automated tuning risks black-box behaviour

What condition makes mechanical AI most useful?

1. Repetitive tasks humans find undesirable

Why: Mechanical AI excels in automating tasks like checkout, cleaning, or delivery, especially when humans dislike them

Which scenario best illustrates the augmentation-replacement cycle?

1. AI begins by helping with routine work and later fully automates it 

Why: This pattern has occurred in manufacturing, service, and emotional contexts

What most clearly differentiates AI from earlier technologies?

1. Its self-learning and autonomous adaptation capabilities

 Why: Unlike traditional IT, AI can learn and adapt autonomously without new programming.

Which factor most strongly influenced the decision to use CSMs for modelling Twitter behaviour?

1. Interpretability of behavioural states

 Why: CSMs were selected for their transparent, interpretable models of user behavior over time.

What is the potential ethical risk of AI-driven price targeting, particularly in the EU?

1. Discrimination based on group characteristics

 Why: The EU considers algorithmic pricing based on group characteristics discriminatory.

How does the CRISP-DM framework support AI implementation in marketing?

It provides a repeatable, iterative process tailored for marketing AI 

 Why:  CRISP-DM is emphasized as a repeatable and flexible process that fits marketing AI projects, adapted by the authors for their use cases.

What best characterises the CRISP-DM framework in practice?

1. A cyclic and iterative methodology allowing feedback loops

 Why: CRISP-DM allows for iteration between phases—especially modeling, evaluation, and data prep.

What is a core reason an image scoring model would require offline learning?

1. Image processing via CNNs is computationally intensive

 Why: CNNs require intensive training on large datasets, making offline learning more practical.

Which learning method best fits a system that learns from feedback after each decision?

1. Reinforcement learning

 Why: Reinforcement learning learns from reward signals after actions, which matches this scenario.

 What kind of learning does reinforcement learning primarily depend on?

1. Feedback from actions taken

 Why: Reinforcement learning learns from real-time trial-and-error feedback.

Which stage of the marketing planning process directly involves identifying anomalies and predicting macro trends using AI?

1. Analyse the current situation

 Why: Need to figure out why

What best describes an ensemble model in AI?

1. A model combining multiple ML models for better predictions

 Why: Ensemble models blend multiple models to improve predictions.

Which type of intelligence is uniquely strong in humans?

1. Contextual and intuitive

 Why: Humans are better at interpreting context and applying intuition, especially in uncertain or emotional domains.

1. What characterizes mechanical intelligence in the AI-HI framework?

1. Performing routine, repetitive tasks with minimal learning

Why: Mechanical intelligence is low-level AI best at routine, repetitive tasks with minimal adaptability.

What is meant by "feeling intelligence" in AI?

1. AI that reacts to emotional data using analytical models

 Why: Feeling AI is not truly emotional. It analytically processes emotional data like tone, sentiment, or facial cues.

Inbound marketing and lead generation

1. Inbound marketing and lead generation

 Why: Albert helped generate leads and significantly increased inbound sales calls.

Why would both logistic regression and CSMs be used in a social media model?

1. To combine static and dynamic data for classification

 Why: Logistic regression captures profile-level variables, while CSMs handle behavioral time-series data.

What is the foundational assumption behind the collaborative AI framework proposed by Huang and Rust (2022)?

1. AI and human intelligence offer complementary strengths

 Why: The framework is built on the premise that AI and HI (human intelligence) each have strengths at different intelligence levels, which makes collaboration productive.

What is the primary function of feature engineering?

1. It converts raw data into more predictive formats

 Why: Feature engineering transforms raw data into features that are more meaningful or predictive for ML models.

What marketing danger is identified when AI is used inappropriately?

1. Misuse of AI can alienate consumers and reduce effectiveness

 Why: Poorly matched AI (e.g., chatbots in emotional contexts can lead to discomfort and backlash.