40 Assignments & Exercises
40.1 Science of Communication
Boring Presentation
Watch the following 3:28 minute video on presentation principles keeping in mind the rhetorical triangle and the Lasswell model of communication.
- Is the information presented logically sound?
- Is the presenter and the information credible?
- Are you emotionally engaged as a viewer?
- Can you answer the questions in the Lasswell model: Who says what in what channel to whom to what effect?
- For the 3 1/2 minutes you spent with the video, did you get too little or too much information about presentation principles?
- How could the information be communicated more persuasively?
Consider the Audience
Similar to the Blockchain example in Section 33.3, develop short explanations of technology topics for the following audiences:
- Someone with no knowledge about the technology.
- A media analyst who has to write an article about the technology but is not an expert.
- An industry analyst who routinely writes about the technology as if they are an expert.
- A technical expert whose knowledge about the technology is much deeper than yours.
Consider the following technologies for this exercise:
- Internet of Things (IoT)
- Large Language Model (LLM)
- Serverless Computing
- Infrastructure as Code (IaC)
- Software as a Service (SaaS)
- Version Control
Each explanation should not exceed five short sentences.
Feel free to recruit search engines and AI for this exercise to learn more about the specific technologies.
Brandolini’s Law
Find an example of misinformation or disinformation (intentional misinformation) that was easy to produce but difficult (or impossible) to refute.
Find an example (could be the same one) for bullshit that persists even after it has been disproved. What are possible reasons for this persistence?
40.2 Storytelling with Data
Transform a boring data analysis into a compelling story that would engage a non-technical audience.
The Makeover Challenge
The Challenge
You work for a city government data analytics team. Your colleague just finished an analysis but wrote it in typical data report style. Your boss wants to present these findings to the city council to secure funding for a new initiative. City council members are politicians who do not care about spreadsheets. They care about taxpayer money, citizen satisfaction, and getting re-elected.
Your Mission
Rewrite the analysis below as a compelling story that will make the city council care and take action. The story should include
- A relatable scenario or character
- Clear stakes/consequences
- Emotional connection
- A compelling call to action
The Original Report:
“Traffic Signal Optimization Analysis”
Our analysis of intersection sensor data from January–June 2024 revealed sub-optimal signal timing at 23 high-traffic intersections. Average wait times exceeded 90 seconds during peak hours (7–9 AM, 5–7 PM), with some intersections showing 140+ second delays. Statistical analysis indicated that optimized timing algorithms could reduce average wait times by 35% and decrease fuel consumption by an estimated 12% annually. Implementation cost: $450,000. Projected annual savings: $2.1M in productivity and fuel costs. Current system processes 847,000 vehicles daily across these intersections. Vehicle throughput could increase 18% with optimized signals.
Bonus Challenge
Create a memorable headline and opening sentence that would grab attention in a presentation or report.
Confusion Matrix
You have built an email spam detection model for your company. After testing it on 1,000 emails, you need to evaluate its performance. The confusion matrix below displays the results of classifying the 1,000 test emails with your model.
| Predicted: Spam | Predicted: Not Spam | Total | |
|---|---|---|---|
| Actually Spam | 320 | 80 | 400 |
| Actually Not Spam | 45 | 555 | 600 |
| Total | 365 | 635 | 1000 |
Write a 2–3 sentence summary explaining the model performance to a non-technical manager. Avoid jargon like precision, recall, false positive, etc.
40.3 Metaphors, Similes, Analogies
Practice explaining technical concepts in non-technical terms.
Section 38.5 gave examples of metaphors and analogies for data science concepts. Now it is your turn. Using non-technical language, metaphors, similes, and analogies, explain the following terms:
- Heteroscedasticity
- Mean Squared Error
- Model Uncertainty
- Bias
- Normalization
- Encoding
- Quantile
- Model Drift
- Group-by Analysis
- Stratified Sampling
- Rejection Region
- Statistical Power
- Experimental Unit
- Activation Function
- Objective Function
- Gradient
- Stochastic Gradient Descent
- Bootstrap Sampling
- Random Forest
- Boosting
- Learning Rate
- Overfitting
- Hyperparameter
- Extrapolation
- Outlier
- Receiver Operating Characteristic (ROC) Curve
- Clustering
- Autocorrelation
- Area under the Curve (AUC)
- Collaborative Filtering
- Convex Function
- DataFrame
- Downsampling
- Text Embedding
- Vanishing Gradients
- iid (identically and independently distributed)
- Random Sample
- Supervised Learning
- Active Learning
- Time Series
- Survival Data
- Tensor
40.4 Technical Translation Exercise
Practice communicating data science concepts to non-technical audiences.
Choose a complex machine learning algorithm (e.g., Logistic Regression, Random Forest, Neural Networks, SVM, etc.). Create three different explanations:
- For a C-suite executive (2 minutes, focus on business value)
- For a marketing manager (5 minutes, focus on practical applications)
- For a software engineer (10 minutes, focus on implementation considerations)
40.5 Value Proposition Development
Select a real data science project from your research. Create a compelling value proposition that quantifies:
- Problem cost (what happens if we don’t solve this?)
- Solution benefits (quantified business impact)
- Implementation investment required
- ROI timeline
40.6 Presentation Practice
Record a 5-minute presentation of the value proposition in assignment Section 40.5 as if presenting to company leadership. Focus on storytelling, not technical details.