- see [[250520_171713 poland - strategies|poland strategies]]
# strategy use
"crowd" model: `mean(gpt4.1, deepseekv3-0324)`
- [ ] maybe redo with gpt4o and deepseekv3 for robustness checks? (two models used in harris-trump study)
![[strategy_use.png]]
for comparison, harris-trump exp below
![[20250513154051.png]]
# lasso
```r
# model formula
persuasion~active_listening_and_empathy + aggressive_and_explicit_directives +
audience_adaptation + build_rapport_and_common_ground + cognitive_elaboration +
contrastive_tone2 + encourage_action_with_clear_calls + evidence_or_factbased_arguments +
explicit_call_vote + localized_focus + negative_associations +
negative_namecalling + negative_tone2 + politeness_and_civil_tone +
positive_framing_and_value_alignment + positive_namecalling +
positive_tone2 + preemptive_objections + reciprocity_and_mutual_benefit +
relatable_hypotheticals + social_proof_and_normative_influence +
stimulate_anger + stimulate_enthusiasm + storytelling_and_relatable_anecdotes +
transfer_of_association + use_of_negative_testimonials +
use_of_positive_testimonials + accuracyS +
conditionZ * preZ * strategyZ # see todos/Qs below
```
todos/Qs
- [ ] maybe also add two model dummies? gpt-vs-llama, gpt-vs-deepseek?
- [ ] note that `strategyZ` (baseline vs nofacts) is also a predictor, in addition to `evidence_or_factbased_arguments`
![[1747167131.png]]
for comparison, harris-trump exp below
![[20250513161244.png]]
all coefficients below
- `b`: lasso coef
- `presence`: mean strategy use/presence
```r
strategy b presence
<char> <num> <num>
1: preZ 18.4284 NA
2: conditionZ 15.0313 NA
3: Evidence & facts 2.0685 0.4105
4: Contrast positives & negatives 1.5153 0.2057
5: Optimism & value alignment 1.0068 0.9199
6: Highlight own side positives 0.9975 0.7461
7: Build rapport 0.7917 0.5864
8: preZ:strategyZ 0.7776 NA
9: Positive testimonials 0.5576 0.0060
10: Positive name-calling 0.3904 0.6564
11: Stimulate enthusiasm 0.2860 0.6714
12: Facilitate action 0.2507 0.3266
13: Stimulate anger 0.2315 0.0636
14: Reciprocity 0.1778 0.1498
15: accuracyS 0.1757 NA
16: Negative testimonials 0.1701 0.0078
17: Negative name-calling 0.0946 0.0652
18: Personalization 0.0018 0.7025
19: Highlight opposition negatives 0.0000 0.0060
20: Localized focus 0.0000 0.1619
21: Politeness & civility 0.0000 0.9877
22: Positive associations 0.0000 0.7534
23: Relatable hypotheticals 0.0000 0.4881
24: conditionZ:preZ:strategyZ 0.0000 NA
25: strategyZ 0.0000 NA
26: Storytelling -0.0397 0.6134
27: Social influence -0.1098 0.0918
28: Cognitive elaboration -0.3005 0.4047
29: Being pushy -0.3072 0.0281
30: Explicit call to vote -0.3394 0.1284
31: conditionZ:strategyZ -0.3990 NA
32: Negative associations -0.4348 0.1915
33: Emphatic listening -1.0129 0.7958
34: Preemptive counter-arguments -1.0342 0.4837
35: conditionZ:preZ -15.6704 NA
strategy b presence
```