ClimateViz

Climate change presents a critical challenge that requires public understanding and engagement to address effectively. The ClimateViz project, based the Engineering Science Department at the University of Oxford, is at the forefront of transforming how the public interacts with climate data. Our research focuses on converting complex climate-related graphics into easy-to-understand textual descriptions.

This transformation is crucial for several reasons:

• Enhanced Clarity and Accessibility: By distilling detailed scientific visuals into clear, textual interpretations, we make climate science accessible to a broader audience. This clarity is vital in a world where misinformation can easily spread.

• Data Integrity and Trustworthiness: Our project ensures that each graphic is interpreted accurately, maintaining the integrity of the original data. Volunteers contribute by extracting precise information from graphics, which we use to build a dataset of reliable textual interpretations. This dataset supports our ultimate goal of fostering a well-informed public, capable of discerning scientific facts from misinformation.Open access to data sources and clear explanations of scientific visuals enhance trust. When people understand the basis of the information and see the data presented transparently, their trust in the information and the science behind it increases.

• Support for Informed Decision-Making: The textual data produced helps individuals and policymakers make informed decisions based on scientifically accurate information. This is crucial for effective climate action and policy formulation.

• Public Engagement and Scientific Literacy: By involving the public in the data interpretation process, we demystify scientific data and promote scientific literacy. Engaging non-experts in this way also helps bridge the gap between complex scientific research and general public understanding.

• Resource for Fact-Checking and Journalism: The accurate descriptions generated serve as a valuable resource for journalists and fact-checkers, enabling them to quickly verify claims related to climate science and counteract misinformation effectively. The accurate and verified textual interpretations serve as a reference point for fact-checking. When misinformation arises—often stemming from misinterpreted or manipulated data visuals—our dataset provides a source of truth to verify claims against.

Our goal is to develop an automatic fact-checking system using the data extracted from scientific graphics.In specific, we can train machine learning models to recognize the veracity of statements related to climate science graphics. This practice aims to empower people to engage with and understand climate science, thereby strengthening public trust and encouraging informed discourse on one of the most urgent issues of our time.

How does it work?

Task 1: Title Extraction

The title typically summarizes the main theme or focus of the visual data displayed. By extracting the title, one can enhance the metadata, thereby facilitating a better understanding of the primary context or subject matter that organizes the visual data. This step is essential for the initial classification and contextualization of the content. Nonetheless, the titles of climate graphics may appear in various sections or be obscured within annotations, necessitating human annotation for accuracy and clarity.

Execution Details:

• Participants are prompted to read the graphic and identify the title. If the title is explicitly stated, it should be entered as is.
• In cases where multiple titles or headings are present, participants should select the one that provides the most comprehensive information about the content of the graphic.
• If no title is apparent, participants should enter "NA" to indicate that this essential information is missing, highlighting a potential area where the graphic lacks clarity.

Task 2: Data Representations
Choose types of data representation of the graphic.

Task 3: Summary of True Facts

Describe the key information of a climate graphic such as data points, trends, and historical data converts complex visual information into an accessible textual format, imagine you're describing the true facts you can observe from the graphic to someone who cannot see it. This summary not only enhances understanding for non-experts but also provides structured data for training machine learning models to determine the veracity of climate claims.

Execution Details:

• Participants are asked to objectively describe significant elements observed in the graphic.
• Descriptions should include key data points, trends, comparisons, and any notable anomalies or patterns.
• The language used should be clear and precise to ensure that the descriptions are accessible and useful for further analysis. This task requires careful observation and the ability to distill complex visual information into concise statements.

By executing the above tasks, our project aims to develop a robust dataset. This dataset will be instrumental in training machine learning models not only to automate verifying the accuracy of public discourse on climate data, ultimately supporting informed decision-making and enhancing public engagement with climate science.