Skip to main content

Influence Tactics Analysis Results

32
Influence Tactics Score
out of 100
68% confidence
Moderate manipulation indicators. Some persuasion patterns present.
Optimized for English content.
Analyzed Content
X (Twitter)

Robert Youssef on X

But here's where it gets weird. A 2024 study from Waseda University tested the SAME approach on older models and found the opposite: GPT-3.5: Accuracy dropped from 60% (polite) to 52% (rude) Llama-2-70B: COLLAPSED from 55% to 28% Same rudeness. Opposite results. What changed?

Posted by Robert Youssef
View original →

Analysis Factors

Confidence
False Dilemmas 2/5
No forced choices; merely questions change without presenting only two extreme options.
Us vs. Them Dynamic 2/5
Mild 'older models' vs. implied 'newer' framing, but no strong us-vs-them dynamics between groups.
Simplistic Narratives 4/5
'Same rudeness. Opposite results' presents binary shift from old to new models as a simple evolution, glossing over nuances.
Timing Coincidence 1/5
Timing appears organic with no correlation to major events; past 72 hours featured unrelated AI news like Meta's nuclear deals and D-Wave acquisition, while studies date to 2024/2025.
Historical Parallels 1/5
No parallels to propaganda; direct reference to peer-reviewed papers (Waseda 2024, Penn State 2025) in organic AI discussions, absent psyop hallmarks.
Financial/Political Gain 2/5
Vague benefits to AI influencers sharing for engagement (e.g., @rryssf_, godofprompt), but no clear political or financial promotion; academic studies from neutral universities.
Bandwagon Effect 1/5
No claims of widespread agreement; focuses on specific study contrasts without implying 'everyone knows' this shift.
Rapid Behavior Shifts 2/5
Mild community discussion on X without urgency or astroturfing; limited posts since Jan 9 allow gradual consideration.
Phrase Repetition 2/5
Similar phrasing in X reposts of original thread, but diverse interpretations across AI community posts; no verbatim coordination across independent outlets.
Logical Fallacies 4/5
Assumes 'SAME approach' implies direct comparability without verifying exact prompt matches or tasks, risking false equivalence.
Authority Overload 2/5
Cites '2024 study from Waseda University' credibly but without specifics on authors or peer-review status.
Cherry-Picked Data 4/5
Selects dramatic drops ('COLLAPSED from 55% to 28%') while omitting full datasets, baselines, or confidence intervals from studies.
Framing Techniques 4/5
Biased emphasis via 'weird', 'COLLAPSED', and question 'What changed?' frames evolution suspiciously rather than neutrally.
Suppression of Dissent 1/5
No labeling of critics; silent on potential counter-studies or limitations.
Context Omission 5/5
Omits study links, full methodology, exact rudeness definitions, and broader context like languages tested or question types, leaving key details out.
Novelty Overuse 3/5
'Here's where it gets weird' and 'Opposite results' highlight surprising contrasts between studies, emphasizing novelty in model behavior shifts.
Emotional Repetition 1/5
No repeated emotional triggers; single instance of emphasis via capitalization like 'COLLAPSED' without looping phrases.
Manufactured Outrage 3/5
Phrasing like 'COLLAPSED from 55% to 28%' dramatizes older model failure, but ties to factual study data rather than baseless anger.
Urgent Action Demands 1/5
No demands for immediate action; content poses a rhetorical question 'What changed?' inviting reflection rather than pressuring response.
Emotional Triggers 3/5
Uses words like 'weird' and 'COLLAPSED' to evoke surprise and concern over unexpected results, subtly triggering curiosity or unease about AI changes without extreme fear or outrage.

Identified Techniques

Doubt Name Calling, Labeling Slogans Appeal to Authority Repetition

What to Watch For

Notice the emotional language used - what concrete facts support these claims?
This content frames an 'us vs. them' narrative. Consider perspectives from 'the other side'.
Key context may be missing. What questions does this content NOT answer?

This content shows some manipulation indicators. Consider the source and verify key claims.

Was this analysis helpful?
Share this analysis
Share on X Facebook LinkedIn
Analyze Something Else