Fast and Slow Decision Making with Data Visualizations

Can the way data are represented influence how quickly and accurately we make decisions?

According to Daniel Kahneman's theory of fast and slow thinking, the brain uses two distinct cognitive processes. By understanding how these processes interact with data visualizations, we can design tools that not only catch attention but also improve decision-making.

Thinking Fast and Slow: How Our Two Mind Systems Affect Data Analysis and Decision Making

Daniel Kahneman in his book "Thinking, Fast and Slow " described two systems of thinking and how these systems impact the human ability to analyze data and make decisions. System 1 (thinking fast) is automatic, intuitive, and emotional, while System 2 (thinking slow) is deliberate and calculated. These two systems constantly interact, influencing our thoughts, judgments, and actions.

System 1: Fast and Intuitive

System 1 relies on heuristics and mental shortcuts to make quick judgments, which can be helpful in familiar situations but can lead to errors in data analysis.?

• This system is prone to biases like the halo effect, where a positive impression of one aspect leads to positive judgments about other unrelated aspects.?
• It is also susceptible to confirmation bias, where individuals favor information that confirms their existing beliefs.?
• It’s lazy and tends to follow offered options. For instance, when presented with a question like "Is James friendly?" without any other information, people tend to agree because System 1 automatically favors the suggested idea.
• It hates numbers. For example, the bat-and-ball problem, where a bat and ball cost $1.10 and the bat is $1 more expensive than the ball, often leads people to incorrectly answer $0.10 for the ball's price because System 1 jumps to a quick conclusion, while careful calculation leads to the correct $0.05 answer.

System 2: Slow and Analytical

System 2 requires more effort and concentration, enabling individuals to analyze data more carefully and make more rational decisions.?

• Cognitive strain, often triggered by complex or unfamiliar information, can activate System 2, leading to more deliberate data analysis. This can be achieved by presenting information in a challenging way, such as using a difficult-to-read font, forcing the mind to pay more attention.

However, even when System 2 is engaged, human decision-making is not always perfectly rational.?

• Prospect theory challenges the idea that individuals always make choices based on pure utility. Instead, emotional factors, such as fear of loss, can heavily influence decisions. For example, people are more likely to gamble to avoid a loss than to secure an equivalent gain, demonstrating the impact of emotions on decision-making even when a rational analysis might suggest otherwise.
• Mental imagery plays a significant role in data analysis and decision-making, often leading to overconfidence in potentially inaccurate mental pictures. People construct coherent mental images to understand situations, but these images can be misleading, especially when data contradicts them. For instance, relying on a mental image of sunny summer weather might lead someone to dress inappropriately for a cold summer day. Reference class forecasting, which involves using historical data instead of general mental images, can help overcome this issue and improve predictions.

Fast and Slow Decision-Making with Visualizations: How Our Two Mind Systems Process Visualizations

How System 1 processes Decision-Making with Visualizations

There are two key ways in which System 1 influences decision-making with visualizations:

• Bottom-up attention
• Visual-spatial biases

Bottom-up Attention

Bottom-Up Attention is stimulus-driven, it's the immediate sensory input we receive. Visualizations often contain distinguished elements that draw viewers' attention involuntarily. Bottom-up attention can have both positive and negative effects on decision-making.

• Positive Impact: Salient features can guide attention toward task-relevant information, leading to improved decision-making. For example, using color to highlight important data points can facilitate better understanding.
• Negative Impact: On the other hand, overly prominent elements can distract viewers from important but less prominent information. This can lead to viewers overlooking relevant data, as demonstrated by the "foreground effect", where individuals focus on visually prominent information while ignoring less salient data, even if it is critical for the task.

Visual-Spatial Biases

Visual-spatial biases are defined as heuristics or mental shortcuts that arise directly from the visual encoding technique. These biases often have a negative impact as they can lead to systematic errors in judgment, often because viewers misinterpret the visual representation of the data.

• One common visual-spatial bias is the containment heuristic, where viewers perceive data points within a visual boundary as more similar than those outside the boundary. This bias can be problematic when visualizations use boundaries to represent continuous data, as viewers may make inaccurate comparisons based on these artificial boundaries.
• Another bias, called deterministic construal error, occurs when viewers misinterpret visualizations of uncertain data as representing deterministic or certain information. For instance, viewers might perceive error bars around a mean temperature prediction as representing the high and low temperatures instead of a range of possible temperatures.

How System 2 processes Decision-Making with Visualizations

System 2 processing can be consciously activated by the viewer through Top-Down attention; or triggered externally by some degree of a Cognitive Fit misalignment.

Cognitive Fit

Cognitive fit refers to the degree of alignment between the visualization, the viewer's mental schema (prior knowledge about the type of visualization), and the task or question at hand.

• Mismatch and Mental Transformations: When there is a mismatch between these components, viewers need to engage in mental transformations to align the visualization with their understanding and the task requirements. These transformations require significant working memory resources.
• Impact on Performance: Poor cognitive fit can lead to slower decision-making and an increased likelihood of errors, especially for individuals with lower working memory capacity.?

Conversely, good cognitive fit results in faster and more accurate decisions because minimal mental transformations are needed.

Top-Down Attention?

Top-Down Attention is goal-driven, influenced by personal motivation. In this case viewers will consciously analyze visualization based on their knowledge.?

Knowledge-driven processing, influenced by both short-term and long-term knowledge, plays a significant role in decision-making with visualizations. It can moderate the impact of visualization affordances and biases, leading to either more accurate or less accurate interpretations depending on the individual's knowledge and its application.

• Beneficial Impact: Viewers can use their knowledge to override misleading visualizations or to focus their attention on relevant features or data.?
• Limited Impact: In some cases, knowledge might not be sufficient to overcome strong visual-spatial biases. Viewers may stick to familiar interpretations even when presented with information that contradicts those interpretations.

Summary?

The dual-process account provides a useful framework for understanding how people make decisions with visualizations. It highlights the importance of considering both automatic, perception-driven processes and more deliberate, knowledge-based processes in visualization design.

• Leverage System 1: Design visualizations that minimize potential misinterpretation due to visual biases and use saliency to guide attention to task-relevant information.
• Facilitate System 2: For complex tasks that require deliberate reasoning, consider providing users with clear instructions and training to enhance cognitive fit and support effective mental transformations.?
• Individual Differences: It’s also important to take into account factors like working memory capacity, graph literacy, and numeracy. These factors can impact how individuals process and interpret visualizations.