An official website of the United States government
Attempting to make sense of a phenomenon or crisis, social media users often share data visualizations and interpretations that can be erroneous or misleading. Prior work has studied how data visualizations can mislead, but do misleading visualizations reach a broad social media audience? And if so, do users amplify or challenge misleading interpretations? To answer these questions, we conducted a mixed-methods analysis of the public's engagement with data visualization posts about COVID-19 on Twitter. Compared to posts with accurate visual insights, our results show that posts with misleading visualizations garner more replies in which the audiences point out nuanced fallacies and caveats in data interpretations. Based on the results of our thematic analysis of engagement, we identify and discuss important opportunities and limitations to effectively leveraging crowdsourced assessments to address data-driven misinformation.
more »
« less
This content will become publicly available on April 25, 2026
Linting is People! Exploring the Potential of Human Computation as a Sociotechnical Linter of Data Visualizations
Traditionally, linters are code analysis tools that help developers by flagging potential issues from syntax and logic errors to enforcing syntactical and stylistic conventions. Recently, linting has been taken as an interface metaphor, allowing it to be extended to more complex inputs, such as visualizations, which demand a broader perspective and alternative approach to evaluation. We explore a further extended consideration of linting inputs, and modes of evaluation, across the puritanical, neutral, and rebellious dimensions. We specifically investigate the potential for leveraging human computation in linting operations through Community Notes—crowd-sourced contextual text snippets aimed at checking and critiquing potentially accurate or misleading content on social media. We demonstrate that human-powered assessments not only identify misleading or error-prone visualizations but that integrating human computation enhances traditional linting by offering social insights. As is required these days, we consider the implications of building linters powered by Artificial Intelligence.
more »
« less
- Award ID(s):
- 2402719
- PAR ID:
- 10618398
- Publisher / Repository:
- ACM
- Date Published:
- ISBN:
- 9798400713958
- Page Range / eLocation ID:
- 1 to 10
- Format(s):
- Medium: X
- Location:
- Yokohama Japan
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Secure multi-party computation has seen substantial performance improvements in recent years and is being increasingly used in commercial products. While a significant amount of work was dedicated to improving its efficiency under standard security models, the threat models do not account for information leakage from the output of secure function evaluation. Quantifying information disclosure about private inputs from observing the function outcome is the subject of this work. Motivated by the City of Boston gender pay gap studies, in this work, we focus on the computation of the average of salaries and quantify information disclosure about private inputs of one or more participants (the target) to an adversary via information-theoretic techniques. We study a number of distributions including log-normal, which is typically used for modeling salaries. We consequently evaluate information disclosure after repeated evaluation of the average function on overlapping inputs, as was done in the Boston gender pay study that ran multiple times, and provide recommendations for using the sum and average functions in secure computation applications. Our goal is to develop mechanisms that lower information disclosure about participants’ inputs to a desired level and provide guidelines for setting up real-world secure evaluation of this function.more » « less
-
Large Language Models (LLMs) have received much recent attention due to their human-level accuracy. While existing works mostly focus on either improving accuracy or testing accuracy robustness, the computation efficiency of LLMs, which is of paramount importance due to often vast generation demands and real-time requirements, has surprisingly received little attention. In this article, we make the first attempt to understand and test potential computation efficiency robustness in state-of-the-art LLMs. By analyzing the working mechanism and implementation of 20,543 public-accessible LLMs, we observe a fundamental property in LLMs that could be manipulated in an adversarial manner to reduce computation efficiency significantly. Our interesting observation is that the output length determines the computation efficiency of LLMs instead of the input, where the output length depends on two factors: an often sufficiently large yet pessimistic pre-configured threshold controlling the max number of iterations and a runtime-generated end of sentence (EOS) token. Our key motivation is to generate test inputs that could sufficiently delay the generation of EOS such that LLMs would have to go through enough iterations to satisfy the pre-configured threshold. We presentLLMEffiChecker, which can work under both white-box setting and black-box setting. In the white-box scenario,LLMEffiCheckerdevelops a gradient-guided technique that searches for a minimal and unnoticeable perturbation at character-level, token-level, and structure-level. In the black-box scenario,LLMEffiCheckeremploys a causal inference-based approach to find critical tokens and similarly applies three levels of imperceptible perturbation to them. Both the white-box and black-box settings effectively delay the appearance of EOS, compelling these inputs to reach the naturally unreachable threshold. To demonstrate the effectiveness ofLLMEffiChecker, we conduct a systematic evaluation on nine publicly available LLMs: Google T5, AllenAI WMT14, Helsinki-NLP translator, Facebook FairSeq, UNICAMP-DL translator, MarianMT, Google FLAN-T5, MBZUAI LaMini-GPT, and Salesforce CodeGen. Experimental results show thatLLMEffiCheckercan increase on average LLMs’ response latency and energy consumption by 325% to 3,244% and 344% to 3,616%, respectively, by perturbing just one character or token in the input sentence. Our case study shows that inputs generated byLLMEffiCheckersignificantly affect the battery power in real-world mobile devices (i.e., drain more than 30 times battery power than normal inputs).more » « less
-
Chart construction errors, such as truncated axes or inexpressive visual encodings, can hinder reading a visualization, or worse, imply misleading facts about the underlying data. These errors can be caught by critical readings of visualizations, but readers must have a high level of data and design literacy and must be paying close attention. To address this issue, we introduce VisuaLint: a technique for surfacing chart construction errors in situ. Inspired by the ubiquitous red wavy underline that indicates spelling mistakes, visualization elements that contain errors (e.g., axes and legends) are sketchily rendered and accompanied by a concise annotation. VisuaLint is unobtrusive — it does not interfere with reading a visualization — and its di- rect display establishes a close mapping between erroneous elements and the expression of error. We demonstrate five examples of VisualLint and present the results of a crowdsourced evaluation (N = 62) of its efficacy. These results contribute an empiri- cal baseline proficiency for recognizing chart construction errors, and indicate near-universal difficulty in error identification. We find that people more reliably identify chart construction errors after being shown examples of VisuaLint, and prefer more verbose explanations for unfamiliar or less obvious flaws.more » « less
-
The growing popularity of interactive time series exploration platforms has made data visualization more accessible to the public. However, the ease of creating polished charts with preloaded data also enables selective information presentation, often resulting in biased or misleading visualizations. Research shows that these tools have been used to spread misinformation, particularly in areas such as public health and economic policies during the COVID-19 pandemic. Post hoc fact-checking may be ineffective because it typically addresses only a portion of misleading posts and comes too late to curb the spread. In this work, we explore using visualization design to counteract cherry-picking, a common tactic in deceptive visualizations. We propose a design space of guardrails—interventions to expose cherry-picking in time-series explorers. Through three crowd-sourced experiments, we demonstrate that guardrails, particularly those superimposing data, can encourage skepticism, though with some limitations. We provide recommendations for developing more effective visualization guardrails.more » « less
