Tip #20 Reading: Multiple-Choice [Multiple Answers] (Task 2): Use a triple phase preview (scan, skim, predict), have a first phase elimination process (based on answers that contradict text, are too broad/narrow, use extreme language without proof

.

1. Pre-Reading Strategies (1–2 mins)
A. Triple-Phase Preview:

1. Scan Question Type: Identify if it tests content (facts) or tone (author's attitude).
2. Skim Options First: Look for:
   • Extreme words (all, never, must → often wrong).
   • Matching phrases that might be "traps."
3. Predict Answers: Based on options, guess which 2-3 might be correct before reading.

B. Targeted Reading:

• Focus on topic sentences (1st/last lines of paragraphs) and contrast words (however, but).
• Underline keywords that match options (e.g., dates, names, opinions).

2. Advanced Elimination Techniques
Step 1: First Pass Elimination

• Delete options that:
  • Contradict the text (even if one word matches).
  • Are too broad/narrow vs. the text’s scope.
  • Use extreme language without proof.

Step 2: Deep-Validate Survivors
For remaining options:

• Exact Matches: Beware! Some repeat text words but distort meaning.
  • Trap Example: Text says "mitigate pollution," option says "eliminate pollution."
• Implied Ideas: Check if logically supported (e.g., text criticizes Policy X → implies Policy X is flawed).

Step 3: Tone Analysis (If Asked)

• Positive Tone: Look for "beneficial," "innovative," "optimistic."
• Negative Tone: "problematic," "criticized," "worsen."
• Neutral Tone: "discusses," "presents," "analyzes."

3. Challenging Model Task (300-word Text)
Text (Excerpt on AI Ethics):
"Artificial Intelligence (AI) promises transformative benefits, from medical diagnostics to climate modeling. However, its deployment raises ethical dilemmas. A 2023 Stanford study revealed that 67% of AI systems exhibit bias against marginalized groups, often due to flawed training data. While some argue that AI’s efficiency justifies its risks, critics warn that unchecked automation could erode human agency. For instance, LinkedIn’s algorithm was found to prioritize male candidates for high-paying roles. Regulatory bodies like the EU now mandate ‘right to explanation’ clauses, requiring AI decisions to be interpretable. Yet, enforcement remains inconsistent globally. Proponents counter that AI, if designed inclusively, can reduce human prejudice—a claim supported by MIT’s 2022 trial where bias-aware algorithms improved hiring diversity by 40%."
Question: Which statements reflect concerns raised in the text?

1. AI consistently improves hiring practices.
2. Most AI systems show bias against marginalized groups.
3. The EU enforces strict global AI regulations.
4. AI’s efficiency alone justifies its ethical risks.
5. Lack of transparency in AI decision-making is a documented issue.

Answer at the end

4. Post-Task Assessment
   Content Accuracy:

• Did you avoid options with matching words but wrong meaning (e.g., "improves hiring" vs. "improved diversity in one trial")?
• Did you select only fully supported answers?

   Tone Consistency:

• If tone was asked, did you ignore your own opinions? (e.g., text is critical of AI risks, not neutral).

   Time Check:

• Spent ≤ 4 mins (reading + answering)?

5. Reflection Prompts

1. Trap Options: Which distractor tricked you? Why? (e.g., "AI consistently improves..." used "improves" from the text but added "consistently").
2. Over-Selection: Did you pick extra options "just in case"? List one you doubted.
3. Vocabulary Gaps: Did terms like "bias-aware algorithms" affect your accuracy?

6. Pro Tips for High Difficulty

• Comparative Claims: Options with "more than," "less likely" require explicit text comparisons.
• Partial Truths: Some options are true in isolation but not text-supported (e.g., "AI aids climate modeling" is true but not a concern).
• Reverse-Verify: For each selected option, find 2 pieces of text evidence.

Need Harder? Try This:
"Which reflect the author’s implicit stance?"

1. AI regulation should prioritize corporate interests.
2. Bias in AI systems is inevitable but manageable.
3. The EU’s approach is insufficient.
4. Human judgment is always superior to AI.
5. Transparency requirements stifle innovation.

Answers: 2, 3 ("Bias-aware algorithms" implies manageability; "enforcement inconsistent" suggests EU’s approach is lacking). 1, 4, 5 are unsupported or extreme.

Final Reminder: Treat this task like a detective game—your job is to prove answers with text evidence, not assumptions! 

Correct Answers:

• 2 ("67% of AI systems exhibit bias").
• 5 ("‘right to explanation’ clauses" implies transparency issues).

Why Others Are Wrong:

• 1: "Consistently" is extreme; MIT’s trial shows improvement but not universality.
• 3: "Global" is false; text says "enforcement remains inconsistent globally."
• 4: Text critiques this view ("critics warn...").