A hypothesis is a specific, testable proposition about a relationship or difference. It keeps analysis focused on evidence that can support or contradict a stakeholder’s theory. section 6.2.1 section 6.3
A practical hypothesis starts with a vague business concern and identifies the data needed to test it. One investigation asks whether coding experience changes how developers use, assess, or trust generative AI tools. section 6.2.1
A results-oriented analysis turns hypotheses into a minimum viable investigation:
- Read the survey questions closely and select variables connected to experience, usefulness, trust, or likely influencing factors. section 6.3
- Choose an initial dataset deliberately; using the latest survey can reduce unnecessary scope when AI-tool usage changed substantially over time. section 6.3
- Explore the proposed relationship with grouped comparisons, crosstabs, and heatmaps before deciding whether a statistical test is warranted. section 6.3 section 7.2.1
Experience and trust in AI outputs
After exploratory heatmaps suggest that more experienced developers trust AI outputs less, formulate the testable hypothesis that continuous years of experience affects trust ratings. section 7.2.1
exp_bins = pd.cut(survey_ai_users["YearsCodePro"],
bins=[-1, 0, 2, 5, 10, 20, 50],
labels=["0", "1-2 years", "3-5 years",
"6-10 years", "11-20 years",
"over 20 years"]
)
survey_ai_users.groupby(exp_bins)["YearsCodePro"].agg(["min", "max"])
Operationalize experience by binning YearsCodePro into ordered ranges, then compare trust responses across those groups. section 7.2.1
