Distinguishing causation and correlation: Causal learning from time-series graphs with trends

Area of Science:

• Cognitive psychology
• Data visualization
• Human-computer interaction

Background:

• Time-series graphs are widely used in science, popular media, and mobile health apps.
• Accurate interpretation of relationships in time-series data is crucial for decision-making.
• Temporal trends can complicate the perception of variable relationships in graphs.

Purpose of the Study:

• To investigate human ability to discern relationships between variables in time-series graphs.
• To identify challenges in interpreting causal relations, especially with temporal trends.
• To explore methods for improving the interpretability of time-series data visualizations.

Main Methods:

• Participants were presented with time-series graphs displaying varying relationships between two variables.
• Experimental conditions included graphs with and without strong temporal trends.
• A dynamic presentation method was tested to assess its impact on interpretation.

Main Results:

• Most participants could distinguish positive from negative relationships, even with significant temporal trends.
• A specific challenge emerged when a positive causal relation was paired with opposing temporal trends (one variable increasing, the other decreasing).
• A simple dynamic presentation significantly improved participants' ability to infer the correct causal relation in difficult cases.

Conclusions:

• People's ability to infer causal relationships from time-series graphs is susceptible to confounding temporal trends.
• Opposing temporal trends pose a significant challenge to accurate causal inference.
• Dynamic visualization techniques offer a promising approach to enhance the interpretability of time-series data and support accurate causal reasoning.