Public Understanding

Dealing with Data

When it comes to processing information, humans and machines each have their own way of doing things. Data resembles a jigsaw puzzle, where each piece is essential to the complete picture. When these pieces are put together, organized, and given context, they have meaning and create a story. Let’s draw parallels from everyday human experience.

Human Analogy

Just as assembling a puzzle involves piecing together fragments to reveal the full picture, humans approach complex data by first analyzing and visualizing smaller parts of the whole. These smaller connected parts unravel the meaning of the complexity and allow themselves to be connected more meaningfully. Each subsequent connection reveals more of the story and expands the knowledge of the whole.

Machine Analogy

Think of machines sorting through puzzle pieces to identify patterns and solve the puzzle efficiently. Machines methodically arrange features, much like categorizing puzzle pieces, which affects their ability to effectively identify patterns and make accurate classifications.

Everyday comparison

In everyday scenarios, humans simplify tasks by breaking them down into manageable steps, much like solving a puzzle piece by piece. Conversely, machines streamline processes by systematically structuring data, much like organizing ingredients for a recipe. Both approaches represent different ways of simplifying complexity, combining human creativity with machine precision to improve understanding and decision-making.

Data Representation and Abstraction in Humans and Machines

Humans and machines approach data representation and abstraction differently, affecting how they process and interpret information in various domains, such as machine learning and data visualization.

Data Representation

In traditional machine learning, structured data representation is essential for effectively feeding information into models. Typically, data is organized in two-dimensional tables where rows represent observations and columns represent features. For tasks requiring target labels, the data table is divided into a feature matrix (containing all features except the target) and a target matrix (containing only the target values). This structured approach helps to efficiently train machine learning algorithms.

Data visualization serves as a powerful tool for humans to understand complex data, make informed decisions, and communicate insights effectively. Graphical representations help humans explore nuances, identify errors, and extract valuable information. While AI can automate certain visualization tasks, human cognition excels at capturing subtleties, biases, ethical considerations, and domain-specific details that machines may struggle with.

Data Abstraction

People use data abstraction to simplify complexity and increase understanding. By filtering out irrelevant details, abstraction allows people to focus on the essential aspects of decision-making and problem solving. For example, in everyday tasks such as baking or dressing, abstraction streamlines processes by presenting simplified concepts rather than overwhelming users with intricate details. These concepts can be as simple as determining the correct order of ingredients to bake a cake, or as complex as using color theory to match different outfits to create something unique.

Machines use data abstraction to solve complex problems like jigsaw puzzles in an efficient way. Instead of worrying about the intricate details of each puzzle piece, they represent the pieces as a set of relevant features, such as shape, color, texture, and spatial relationships. This abstraction allows the machine to focus on the essential features needed to solve the puzzle, rather than getting bogged down in the fine-grained details. The machine can then calculate the “distance” between different piece features to determine which ones are likely to fit together, and use this distance-based matching to assemble the puzzle. Through this process of abstraction, feature representation, and distance-based matching, machines can solve complex puzzles without needing to understand the full complexity of each piece.

Human-Machine Collaboration

The divergence between human and machine use of data representation and abstraction stems from their different goals and capabilities. Humans use abstraction to reduce cognitive load, aiding comprehension and task performance by focusing on relevant information. In contrast, machines rely on abstraction to increase efficiency in programming and computing, simplifying tasks through predefined structures created by programmers.

While both humans and machines use data representation and abstraction to simplify and manage complexity, their approaches play to their unique strengths: human cognition excels at understanding context and adapting to new situations, while machines excel at efficiently processing large amounts of structured data within predefined frameworks. By combining the strengths of humans and machines, hidden insights can be uncovered from complex data sets to enable better decision-making.

What We Do

The group is dedicated to developing computational solutions that merge visualization techniques with Artificial Intelligence (AI). Their primary goal is to enhance understanding and control of AI models and data analytics through innovative visualization approaches. This group aims to address challenges like uncertainty, accessibility, and bias in visualized data by leveraging AI methods.