Skip to content

Latest commit

 

History

History
50 lines (43 loc) · 7.4 KB

File metadata and controls

50 lines (43 loc) · 7.4 KB

Prolific Survey Participant Demographics

The age, subject(s) of study, education level, and whether individuals are native English speakers is presented in the table below. One point which stands out is that most participants listed multiple subjects. In some cases these were similar subjects or superdomains of the same subject, such as Chemistry and Biochemistry or Computer Science and Engineering, while in other cases the groupings suggested an interdisciplinary program of study, as in the case of one sociology participant whose reported subjects are "Art and/or design, Communication and/or Media, Computer Science, Fashion and textiles, Sociology". Additionally, while all Computer Science participants had completed only an undergraduate degree, for the other domains two out of three participants had completed at least a graduate degree.

Age Subject(s) Education Level Native English
Chemistry
Participant 1 32 Engineering, Science, Chemistry Graduate
Participant 2 33 Chemistry, Biochemistry (Molecular and Cellular) Undergraduate
Participant 3 24 Chemistry Doctorate
Computer Science
Participant 1 24 Engineering, Economics, Business, Computer Science, Marketing Undergraduate
Participant 2 26 Computing (IT), Engineering, Mathematics, Economics, Computer Science Undergraduate
Participant 3 33 Computer Science, Computing (IT), Engineering, Mathematics, Physics Undergraduate
Earth Science
Participant 1 30 Earth, Environment, or Climate Sciences Undergraduate
Participant 2 29 Earth, Environment, or Climate Sciences, Geography, Science Graduate
Participant 3 28 Earth, Environment, or Climate Sciences, Science Graduate
Linguistics
Participant 1 24 Education, English Language, English Literature, Languages Graduate
Participant 2 23 Business, Languages, Management, Communication and/or Media Graduate
Participant 3 24 Languages Undergraduate
Sociology
Participant 1 30 Sociology, Economics, Philosophy Graduate
Participant 2 24 Art and/or design, Communication and/or Media, Computer Science, Fashion and textiles, Sociology Undergraduate
Participant 3 22 Sociology Graduate

Table 1: Demographic information about study participants. All information is self-reported. Education level refers to highest level of education completed.

Category Relevancy Correctness Completeness Informativeness Integration Cohesion Coherence Readability Conciseness Avg
Chemistry
GPT-4 (Human/Auto) 4.61/5.00 4.38/5.00 4.11/4.66 4.27/5.00 4.05/5.00 3.72/5.00 3.88/5.00 4.22/5.00 4.22/4.00 4.16/4.85
Mistral (Human/Auto) 4.38/4.50 4.27/4.00 4.00/3.66 4.44/4.33 4.05/4.66 3.88/4.83 3.94/4.66 4.27/5.00 3.50/4.00 4.08/4.40
Computer Science
GPT-4 (Human/Auto) 4.66/5.00 4.50/5.00 4.05/4.50 3.88/5.00 4.38/5.00 4.27/5.00 4.27/5.00 4.38/5.00 3.83/3.66 4.24/4.79
Mistral (Human/Auto) 4.38/3.16 4.33/3.33 3.22/3.16 3.22/3.33 3.44/3.16 3.66/3.16 3.83/3.16 4.27/3.83 3.61/2.83 3.77/3.24
Earth Science
GPT-4 (Human/Auto) 4.66/5.00 4.33/4.83 4.22/4.50 4.61/5.00 4.38/5.00 4.55/5.00 4.44/5.00 4.55/4.83 4.72/3.66 4.49/4.75
Mistral (Human/Auto) 4.38/3.66 4.22/3.66 4.16/3.16 4.22/3.66 4.16/4.16 4.11/4.16 4.22/4.00 4.61/4.50 4.44/3.50 4.27/3.83
Linguistics
GPT-4 (Human/Auto) 3.77/5.00 4.11/4.83 3.88/4.50 3.94/4.83 3.72/5.00 3.61/5.00 3.61/5.00 3.94/5.00 3.88/4.00 3.82/4.79
Mistral (Human/Auto) 3.66/4.50 3.38/4.66 3.77/4.00 3.66/4.66 3.50/5.00 3.44/5.00 3.27/4.83 3.72/5.00 3.55/4.16 3.55/4.64
Sociology
GPT-4 (Human/Auto) 4.16/5.00 3.83/5.00 3.55/4.50 4.16/5.00 3.55/5.00 3.72/5.00 3.88/5.00 3.27/5.00 3.77/4.00 3.76/4.83
Mistral (Human/Auto) 3.50/4.66 3.16/3.66 2.33/3.66 3.05/4.33 2.83/4.66 3.38/4.66 3.49/4.66 3.55/4.83 3.72/4.00 3.22/4.35

Table 2: Average human and automatic (LLM) evaluation scores by characteristic. For each domain/characteristic, the human scores are an average of 18 judgements (6 syntheses (2 samples x 3 synthesis types) x 3 participants) while the auto scores are an average of 6 judgements (6 syntheses (2 samples x 3 synthesis types) x 1 LLM evaluation).