About

Arvind Satyanarayan is an Associate Professor in the Department of Electrical Engineering and Computer Science at MIT. His research focuses on human-centered data visualization, developing interactive visualizations that amplify human creativity and cognition. He is involved in projects that explore how design elements of data visualizations influence viewers’ assumptions about information sources and trustworthiness, as well as creating tools that turn everyday objects into animated displays without electronics. His work aims to balance automation and human agency, contributing to the fields of data visualization and human-computer interaction.

Research topics

• Computer science
• Human–computer interaction
• Artificial intelligence
• Programming language
• Data science

Selected publications

• Something In Between Formal Spec and Informal Representation

  2026-01-01

  articleOpen accessSenior author
  PublisherOA PDF

• Creo: From One-Shot Image Generation to Progressive, Co-Creative Ideation

  ArXiv.org · 2026-04-15

  articleOpen accessSenior author

  Text-to-image (T2I) systems enable rapid generation of high-fidelity imagery but are misaligned with how visual ideas develop. T2I systems generate outputs that make implicit visual decisions on behalf of the user, often introduce fine-grained details that can anchor users prematurely and limit their ability to keep options open early on, and cause unintended changes during editing that are difficult to correct and reduce users' sense of control. To address these concerns, we present Creo, a multi-stage T2I system that scaffolds image generation by progressing from rough sketches to high-resolution outputs, exposing intermediary abstractions where users can make incremental changes. Sketch-like abstractions invite user editing and allow users to keep design options open when ideas are still forming due to their provisional nature. Each stage in Creo can be modified with manual changes and AI-assisted operations, enabling fine-grained, step-wise control through a locking mechanism that preserves prior decisions so subsequent edits affect only specified regions or attributes. Users remain in the loop, making and verifying decisions across stages, while the system applies diffs instead of regenerating full images, reducing drift as fidelity increases. A comparative study with a one-shot baseline shows that participants felt stronger ownership over Creo outputs, as they were able to trace their decisions in building up the image. Furthermore, embedding-based analysis indicates that Creo outputs are less homogeneous than one-shot results. These findings suggest that multi-stage generation, combined with intermediate control and decision locking, is a key design principle for improving controllability, user agency, creativity, and output diversity in generative systems.

  PublisherOA PDF

• Belidor: A Specification Language for Operationalizing Structural Analogies Between User Interfaces

  2026-04-13 · 1 citations

  article
  PublisherDOI

• Creo: From One-Shot Image Generation to Progressive, Co-Creative Ideation

  arXiv (Cornell University) · 2026-04-15

  preprintOpen accessSenior author

  Text-to-image (T2I) systems enable rapid generation of high-fidelity imagery but are misaligned with how visual ideas develop. T2I systems generate outputs that make implicit visual decisions on behalf of the user, often introduce fine-grained details that can anchor users prematurely and limit their ability to keep options open early on, and cause unintended changes during editing that are difficult to correct and reduce users' sense of control. To address these concerns, we present Creo, a multi-stage T2I system that scaffolds image generation by progressing from rough sketches to high-resolution outputs, exposing intermediary abstractions where users can make incremental changes. Sketch-like abstractions invite user editing and allow users to keep design options open when ideas are still forming due to their provisional nature. Each stage in Creo can be modified with manual changes and AI-assisted operations, enabling fine-grained, step-wise control through a locking mechanism that preserves prior decisions so subsequent edits affect only specified regions or attributes. Users remain in the loop, making and verifying decisions across stages, while the system applies diffs instead of regenerating full images, reducing drift as fidelity increases. A comparative study with a one-shot baseline shows that participants felt stronger ownership over Creo outputs, as they were able to trace their decisions in building up the image. Furthermore, embedding-based analysis indicates that Creo outputs are less homogeneous than one-shot results. These findings suggest that multi-stage generation, combined with intermediate control and decision locking, is a key design principle for improving controllability, user agency, creativity, and output diversity in generative systems.

  PublisherDOI

• Pluto: Authoring Semantically Aligned Text and Charts for Data-Driven Communication

  2025-03-19 · 2 citations

  articleOpen accessSenior author
  PublisherDOI

• Toward cultural interpretability: A linguistic anthropological framework for describing and evaluating large language models

  Big Data & Society · 2025-01-29 · 14 citations

  articleOpen accessSenior author

  This article proposes a new integration of linguistic anthropology and machine learning (ML) around convergent interests in both the underpinnings of language and making language technologies more socially responsible. While linguistic anthropology focuses on interpreting the cultural basis for human language use, the ML field of interpretability is concerned with uncovering the patterns that Large Language Models (LLMs) learn from human verbal behavior. Through the analysis of a conversation between a human user and an LLM-powered chatbot, we demonstrate the theoretical feasibility of a new, conjoint field of inquiry, cultural interpretability (CI). By focusing attention on the communicative competence involved in the way human users and AI chatbots coproduce meaning in the articulatory interface of human-computer interaction, CI emphasizes how the dynamic relationship between language and culture makes contextually sensitive, open-ended conversation possible. We suggest that, by examining how LLMs internally “represent” relationships between language and culture, CI can: (1) provide insight into long-standing linguistic anthropological questions about the patterning of those relationships; and (2) aid model developers and interface designers in improving value alignment between language models and stylistically diverse speakers and culturally diverse speech communities. Our discussion proposes three critical research axes: relativity, variation, and indexicality.

  PublisherDOI

• Quantifying Visualization Vibes: Measuring Socio-Indexicality at Scale

  ArXiv.org · 2025-08-09

  preprintOpen accessSenior author

  What impressions might readers form with visualizations that go beyond the data they encode? In this paper, we build on recent work that demonstrates the socio-indexical function of visualization, showing that visualizations communicate more than the data they explicitly encode. Bridging this with prior work examining public discourse about visualizations, we contribute an analytic framework for describing inferences about an artifact's social provenance. Via a series of attribution-elicitation surveys, we offer descriptive evidence that these social inferences: (1) can be studied asynchronously, (2) are not unique to a particular sociocultural group or a function of limited data literacy, and (3) may influence assessments of trust. Further, we demonstrate (4) how design features act in concert with the topic and underlying messages of an artifact's data to give rise to such 'beyond-data' readings. We conclude by discussing the design and research implications of inferences about social provenance, and why we believe broadening the scope of research on human factors in visualization to include sociocultural phenomena can yield actionable design recommendations to address urgent challenges in public data communication.

  PublisherOA PDFDOI

• Tactile Vega-Lite: Rapidly Prototyping Tactile Charts with Smart Defaults

  2025-04-24 · 7 citations

  articleOpen access

  CHI ’25, Yokohama, Japan

  PublisherDOI

• GoFish: A Grammar of More Graphics!

  IEEE Transactions on Visualization and Computer Graphics · 2025-12-03 · 2 citations

  articleSenior author

  Visualization grammars from ggplot2 to Vega-Lite are based on the Grammar of Graphics (GoG), our most comprehensive formal theory of visualization. The GoG helped expand the expressive gamut of visualization by moving beyond fixed chart types and towards a design space of composable operators. Yet, the resultant design space has surprising limitations, inconsistencies, and cliffs - even seemingly simple charts like mosaics, waffles, and ribbons fall out of scope of most GoG implementations. To author such charts, visualization designers must either rely on overburdened grammar developers to implement purpose-built mark types (thus reintroducing the issues of typologies) or drop to lower-level frameworks. In response, we present GoFish: a declarative visualization grammar that formalizes Gestalt principles (e.g., uniform spacing, containment, and connection) that have heretofore been complected in GoG constructs. These graphical operators achieve greater expressive power than their predecessors by enabling recursive composition: they can be nested and overlapped arbitrarily. Through a diverse example gallery, we demonstrate how graphical operators free users to arrange shapes in many different ways while retaining the benefits of high-level grammars like scale resolution and coordinate transform management. Recursive composition naturally yields an infinite design space that blurs the boundary between an expressive, low-level grammar and a concise, high-level one. In doing so, we point towards an updated theory of visualization, one that is open to an innumerable space of graphic representations instead of limited to a fixed set of "good" designs.

  PublisherDOI

• Abstraction Alignment: Comparing Model-Learned and Human-Encoded Conceptual Relationships

  2025-04-24 · 1 citations

  articleOpen accessSenior author

  CHI ’25, Yokohama, Japan

  PublisherDOI

Recent grants

• CAREER: Effective Interaction Design for Data Visualization

  NSF · $547k · 2020–2026

• III: Large: Collaborative Research: Analysis Engineering for Robust End-to-End Data Science

  NSF · $729k · 2019–2025

Frequent coauthors

• Jeffrey Heer

  University of Washington

  19 shared

• Jonathan Zong

  Massachusetts Institute of Technology

  15 shared

• Angie Boggust

  11 shared

• Hendrik Strobelt

  9 shared

• Kanit Wongsuphasawat

  9 shared

• Rupayan Neogy

  Massachusetts Institute of Technology

  8 shared

• David Sontag

  Broad Institute

  7 shared

• Alan Lundgard

  Massachusetts Institute of Technology

  7 shared

Labs

• MIT EECS Communication LabPI

Education

• Ph.D., Computer Science

  Stanford University

  2017

• M.S., Computer Science

  Stanford University

  2014

• B.S., Computer Science and Engineering

  University of California San Diego

  2011