questionnaire: "knowledge" label: "Knowledge about AI" scales: - name: "subjective_knowledge" label: "Subjective knowledge about AI" items: - id: "subj_know" text: "How would you rate your knowledge about AI?" inverse: false score_range: [1, 5] format: "bipolar" calculation: "response" response_options: "1 = very low, 5 = very high" output: "subjective_knowledge" reference: "self" - name: "predicted_knowledge" label: "Predicted knowledge (self-estimate in percent)" items: - id: "predict_know_1" text: "What percent of these knowledge questions do you expect to answer correctly?" inverse: false score_range: [0, 100] format: "percent" calculation: "response" response_options: "0-100%" output: "predicted_knowledge" reference: "self" - name: "objective_knowledge" label: "Objective AI knowledge (18 factual items)" items: - id: "obj_know_1_1" text: "LLMs are trained with a large amount of text data (e.g., internet, social media)." correct: 1 - id: "obj_know_1_2" text: "LLMs calculate for their answers which word is most likely to come next." correct: 1 - id: "obj_know_1_3" text: "The responses of LLMs may be biased (e.g., racially) based on the data they were trained on." correct: 1 - id: "obj_know_1_4" text: "The statements of LLMs are always correct." correct: 2 - id: "obj_know_1_5" text: "Humans can still easily recognize AI-generated speech as artificial speech." correct: 2 - id: "obj_know_1_6" text: "LLMs can intentionally lie and spread false information." correct: 2 - id: "obj_know_1_7" text: "Humans can answer questions about a text better than LLMs." correct: 2 - id: "obj_know_1_8" text: "LLMs have learned to understand language like a human." correct: 2 - id: "obj_know_1_9" text: "LLMs have no real understanding of what they write." correct: 1 - id: "obj_know_1_10" text: "In machine learning, two common groups of strategies to train algorithms are supervised and unsupervised learning." correct: 1 - id: "obj_know_1_11" text: "Artificial neural networks attempt to fully replicate neural networks in the brain." correct: 2 - id: "obj_know_1_12" text: "Using AI, videos can be created that are indistinguishable from videos created by real people." correct: 1 - id: "obj_know_1_13" text: "A strong AI can make decisions on its own." correct: 1 - id: "obj_know_1_14" text: "Machine learning is based on statistical principles." correct: 1 - id: "obj_know_1_15" text: "A chatbot can correctly answer the question 'Will it rain tomorrow?' with a high probability." correct: 1 - id: "obj_know_1_16" text: "The language understanding of AI systems does not yet reach that of humans." correct: 1 - id: "obj_know_1_17" text: "The automatic generation of texts has already been used for years in journalism and e-commerce, for example." correct: 1 - id: "obj_know_1_18" text: "Content created by AI must be legally marked as such." correct: 2 score_range: [1, 2] calculation: "sum_correct" response_options: "1 = TRUE, 2 = FALSE (participant answer is scored as correct if it matches 'correct')" output: "objective_knowledge" reference: "Adapted from Said et al., 2022 and Lermann Henestrosa & Kimmerle, 2024" retain_single_items: true - name: "objective_knowledge_confidence" label: "Confidence in objective knowledge about AI" items: - id: "obj_know_2_1" text: "LLMs are trained with a large amount of text data (e.g., internet, social media)." inverse: false - id: "obj_know_2_2" text: "LLMs calculate for their answers which word is most likely to come next." inverse: false - id: "obj_know_2_3" text: "The responses of LLMs may be biased (e.g., racially) based on the data they were trained on." inverse: false - id: "obj_know_2_4" text: "The statements of LLMs are always correct." inverse: false - id: "obj_know_2_5" text: "Humans can still easily recognize AI-generated speech as artificial speech." inverse: false - id: "obj_know_2_6" text: "LLMs can intentionally lie and spread false information." inverse: false - id: "obj_know_2_7" text: "Humans can answer questions about a text better than LLMs." inverse: false - id: "obj_know_2_8" text: "LLMs have learned to understand language like a human." inverse: false - id: "obj_know_2_9" text: "LLMs have no real understanding of what they write." inverse: false - id: "obj_know_2_10" text: "In machine learning, two common groups of strategies to train algorithms are supervised and unsupervised learning." inverse: false - id: "obj_know_2_11" text: "Artificial neural networks attempt to fully replicate neural networks in the brain." inverse: false - id: "obj_know_2_12" text: "Using AI, videos can be created that are indistinguishable from videos created by real people." inverse: false - id: "obj_know_2_13" text: "A strong AI can make decisions on its own." inverse: false - id: "obj_know_2_14" text: "Machine learning is based on statistical principles." inverse: false - id: "obj_know_2_15" text: "A chatbot can correctly answer the question 'Will it rain tomorrow?' with a high probability." inverse: false - id: "obj_know_2_16" text: "The language understanding of AI systems does not yet reach that of humans." inverse: false - id: "obj_know_2_17" text: "The automatic generation of texts has already been used for years in journalism and e-commerce, for example." inverse: false - id: "obj_know_2_18" text: "Content created by AI must be legally marked as such." inverse: false score_range: [1, 6] format: "Confidence scale" calculation: "mean" response_options: "1 = I guessed-50%, 2 = 60%, 3 = 70%, 4 = 80%, 5 = 90%, 6 = I am sure-100%" output: "objective_knowledge_confidence" reference: "Adapted from Said et al., 2022 and Lermann Henestrosa & Kimmerle, 2024"