How Do Analysts Understand and Verify AI-Assisted Data Analyses?
Open accessThe DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index.
Abstract
Data analysis is challenging as it requires synthesizing domain knowledge, statistical expertise, and programming skills. Assistants powered by large language models (LLMs), such as ChatGPT, can assist analysts by translating natural language instructions into code. However, AI-assistant responses and analysis code can be misaligned with the analyst’s intent or be seemingly correct but lead to incorrect conclusions. Therefore, validating AI assistance is crucial and challenging. Here, we explore how analysts understand and verify the correctness of AI-generated analyses. To observe analysts in diverse verification approaches, we develop a design probe equipped with natural language explanations, code, visualizations, and interactive data tables with common data operations. Through a qualitative user study (n=22) using this probe, we uncover common behaviors within verification workflows and how analysts’ programming, analysis, and tool backgrounds reflect these behaviors. Additionally, we provide recommendations for analysts and highlight opportunities for designers to improve future AI-assistant experiences.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Estimating the reproducibility of psychological science
Open Science Collaboration (2015) · Science · doi:10.1126/science.aac4716 · Cited by 8,515 · cited 1×
“Moreover, because conclusions from data analyses often inform high-impact decisions in science [4, 8], business [43, 54, 55], and government [36, 100], it is important for users of AI-based analysis assistants to critically reflect on the assistant’s outputs [87].”
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445922 · Cited by 6,988 · cited 2×
“1) [48, 79, 137] or make seemingly correct erroneous outputs [13, 47].”
Survey of Hallucination in Natural Language Generation
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023) · ACM Computing Surveys · doi:10.1145/3571730 · Cited by 5,025 · cited 2×
“1) [41, 63, 110] or make seemingly correct erroneous outputs [11, 40].”
Survey of Hallucination in Natural Language Generation
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y., Chen, D., Dai, W., et al. (2022) · arXiv · doi:10.48550/arxiv.2202.03629 · Cited by 4,981
1,500 scientists lift the lid on reproducibility
Baker, M. (2016) · Nature · doi:10.1038/533452a · Cited by 4,209
Generative Agents: Interactive Simulacra of Human Behavior
Park, J. S., O'Brien, J., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023) · Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology · doi:10.1145/3586183.3606763 · Cited by 2,187
Generative Agents: Interactive Simulacra of Human Behavior
Park, J. S., O'Brien, J. C., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023) · arXiv · doi:10.48550/arxiv.2304.03442 · Cited by 2,173
Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems
(2020) · doi:10.1145/3313831 · Cited by 1,951
Complacency and Bias in Human Use of Automation: An Attentional Integration
Parasuraman, R., & Manzey, D. H. (2010) · Human Factors: The Journal of the Human Factors and Ergonomics Society · doi:10.1177/0018720810376055 · Cited by 1,710 · cited 1×
“This can be challenging as existing AI assistants lack verification support [106, 112] and even experts can be susceptible to automation bias and overre-liance [100, 132].”
Luger, E., & Sellen, A. (2016) · Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems · doi:10.1145/2858036.2858288 · Cited by 1,176 · cited 1×
“Likewise, since AI-generated analyses can involve sequences of data operations, it is pertinent to create test cases for both the final output and intermediate steps.”
Questioning the AI: Informing Design Practices for Explainable AI User Experiences
Liao, Q. V., Gruen, D., & Miller, S. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376590 · Cited by 925
Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems
(2023) · doi:10.1145/3544548 · Cited by 735
Vaithilingam, P., Zhang, T., & Glassman, E. L. (2022) · CHI Conference on Human Factors in Computing Systems Extended Abstracts · doi:10.1145/3491101.3519665 · Cited by 660
The state of the art in end-user software engineering
Ko, A. J., Abraham, R., Beckwith, L., Blackwell, A., Burnett, M., Erwig, M., Scaffidi, C., Lawrance, J., Lieberman, H., Myers, B., et al. (2011) · ACM Computing Surveys · doi:10.1145/1922649.1922658 · Cited by 633
Grounded Copilot: How Programmers Interact with Code-Generating Models
Barke, S., James, M. B., & Polikarpova, N. (2023) · Proceedings of the ACM on Programming Languages · doi:10.1145/3586030 · Cited by 436 · cited 4×
“While existing work has studied how programmers validate AI-written code for general programming tasks [9], our work observes how data an-alysts’ sensemaking behaviors interplay with their verification of AI-generated analyses.”
InfluentialDeVito, M. A., Birnholtz, J., Hancock, J. T., French, M., & Liu, S. (2018) · Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3173574.3173694 · Cited by 244 · cited 1×
“People develop diverse perceptions and levels of trust in AI systems [28, 39, 51, 64], leading to distinct behaviors when evaluating AI-generated analyses.”
Help Me Help the AI": Understanding How Explainability Can Support Human-AI Interaction
Kim, S. S. Y., Watkins, E. A., Russakovsky, O., Fong, R., & Monroy-Hernández, A. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581001 · Cited by 233
Resilient Chatbots: Repair Strategy Preferences for Conversational Breakdowns
Ashktorab, Z., Jain, M., Liao, Q. V., & Weisz, J. D. (2019) · Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300484 · Cited by 224 · cited 1×
“Works studying abstraction matching [85], prompt writing [139], and repair of AI-generated programs [6, 94] are related but distinct.”
Investigating Explainability of Generative AI for Code through Scenario-based Design
Sun, J., Liao, Q. V., Muller, M., Agarwal, M., Houde, S., Talamadupula, K., & Weisz, J. D. (2022) · 27th International Conference on Intelligent User Interfaces · doi:10.1145/3490099.3511119 · Cited by 219 · cited 5×
“In addition, because understanding the sequence of data manipulations is vital to understanding the output [105, 136], the probe includes descriptions of any intermediate data computed as a result of an atomic data wrangling operation (Fig.”
InfluentialProductivity Assessment of Neural Code Completion
Ziegler, A., Kalliamvakou, E., Li, X. A., Rice, A., Rifkin, D., Simister, S., Sittampalam, G., & Aftandilian, E. (2022) · Proceedings of the 6th ACM SIGPLAN International Symposium on Machine Programming · doi:10.1145/3520312.3534864 · Cited by 200
Productivity Assessment of Neural Code Completion
Ziegler, A., Kalliamvakou, E., Simister, S., Sittampalam, G., Li, A., Rice, A., Rifkin, D., & Aftandilian, E. (2022) · arXiv · doi:10.48550/arxiv.2205.06537 · Cited by 200
Data scientists in software teams: State of the art and challenges
Kim, M., Zimmermann, T., DeLine, R., & Begel, A. (2018) · IEEE Transactions on Software Engineering · doi:10.1109/tse.2017.2754374 · Cited by 199
Conceptual Metaphors Impact Perceptions of Human-AI Collaboration
Khadpe, P., Krishna, R., Fei-Fei, L., Hancock, J. T., & Bernstein, M. S. (2020) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3415234 · Cited by 183 · cited 1×
“People develop diverse perceptions and levels of trust in AI systems [28, 39, 51, 64], leading to distinct behaviors when evaluating AI-generated analyses.”
Complacency and Automation Bias in the Use of Imperfect Automation
Wickens, C. D., Clegg, B. A., Vieane, A. Z., & Sebok, A. L. (2015) · Human Factors: The Journal of the Human Factors and Ergonomics Society · doi:10.1177/0018720815581940 · Cited by 176 · cited 1×
“This can be challenging as existing AI assistants lack verification support [106, 112] and even experts can be susceptible to automation bias and overre-liance [100, 132].”
Managing Messes in Computational Notebooks
Head, A., Hohman, F., Barik, T., Drucker, S. M., & DeLine, R. (2019) · Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300500 · Cited by 165 · cited 2×
“During this sensemaking, analysts often need help managing, tracking, and comparing the iterative and messy nature of their computations [43, 59, 62, 131].”
InfluentialSelf-citationIn-IDE Code Generation from Natural Language: Promise and Challenges
Xu, F. F., Vasilescu, B., & Neubig, G. (2022) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3487569 · Cited by 154 · cited 1×
“1) [48, 79, 137] or make seemingly correct erroneous outputs [13, 47].”
In-IDE Code Generation from Natural Language: Promise and Challenges
Xu, F. F., Vasilescu, B., & Neubig, G. (2021) · arXiv · doi:10.48550/arxiv.2101.11149 · Cited by 153
Exploring exploratory programming
Beth Kery, M., & Myers, B. A. (2017) · 2017 IEEE Symposium on Visual Languages and Human-Centric Computing (VL/HCC) · doi:10.1109/vlhcc.2017.8103446 · Cited by 123 · cited 1×
“Compared to software engineering, data analysis is more exploratory and iterative [49, 70, 85, 90] as code is not the deliverable but the outcome of the analysis [29, 43].”
MethodLiu, M. X., Sarkar, A., Negreanu, C., Zorn, B., Williams, J., Toronto, N., & Gordon, A. D. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3580817 · Cited by 119 · cited 7×
“1 right) [85].”
InfluentialJiang, E., Toh, E., Molina, A., Olson, K., Kayacik, C., Donsbach, A., Cai, C. J., & Terry, M. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3501870 · Cited by 79 · cited 1×
“1) [48, 79, 137] or make seemingly correct erroneous outputs [13, 47].”
Rethinking Explainability as a Dialogue: A Practitioner's Perspective
Lakkaraju, H., Slack, D., Chen, Y., Tan, C., & Singh, S. (2022) · arXiv · doi:10.48550/arxiv.2202.01875 · Cited by 62
Jung, J.-Y., Qiu, S., Bozzon, A., & Gadiraju, U. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3517653 · Cited by 43 · cited 1×
“One option is to incorporate customized feature-rich artifacts (e.g., those in Datamations or DITL [105, 128]) and verification interfaces (e.g., incorporating visualization tools like Voyager [133, 134] or PygWalker [53] for verificaiton).”
Grounded Copilot: How Programmers Interact with Code-Generating Models
Barke, S., James, M. B., & Polikarpova, N. (2022) · arXiv · doi:10.48550/arxiv.2206.15000 · Cited by 23 · cited 4×
“While existing work has studied how programmers validate AI-written code for general programming tasks [9], our work observes how data an-alysts’ sensemaking behaviors interplay with their verification of AI-generated analyses.”
Influential
Cited by
Ironies of Generative AI: Understanding and Mitigating Productivity Loss in Human-AI Interaction
Simkute, A., Tankelevitch, L., Kewenig, V., Scott, A. E., Sellen, A., & Rintel, S. (2024) · International Journal of Human–Computer Interaction · doi:10.1080/10447318.2024.2405782 · Cited by 89
“Otherwise, during this stage, suggestions should carefully align with users’ flow (Gu et al., 2023), in line with ecological interface design.”
InfluentialXie, L., Zheng, C., Xia, H., Qu, H., & Zhu-Tian, C. (2024) · Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology · doi:10.1145/3654777.3676374 · Cited by 52
Beyond the Chat: Executable and Verifiable Text-Editing with LLMs
Laban, P., Vig, J., Hearst, M., Xiong, C., & Wu, C.-S. (2024) · Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology · doi:10.1145/3654777.3676419 · Cited by 31
“In turn, recent work has included these methodologies into systems, allowing users to receive feedback on the factuality of generated text, allowing users to handle AI-generated text with more caution [27, 35, 37, 38, 60, 77].”
Drosos, I., Sarkar, A., Xu, X., Negreanu, C., Rintel, S., & Tankelevitch, L. (2024) · Proceedings of the 3rd Annual Meeting of the Symposium on Human-Computer Interaction for Work · doi:10.1145/3663384.3663389 · Cited by 30
Li, H., Wang, Y., Liao, Q. V., & Qu, H. (2025) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2025.3552017 · Cited by 25
Metadata
- Authors
- Gu, K., Shang, R., Althoff, T., Wang, C., & Drucker, S. M.
- Venue
- Proceedings of the CHI Conference on Human Factors in Computing Systems
- Publisher
- ACM
- Published
- 2024-05-11
- Pages
- 1-22
- OpenAlex
- W4396832076
- Cited by
- 37
- References
- 33
- FWCI
- 8.08
- Type
- conference paper
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.