3DPFIX: Improving Remote Novices' 3D Printing Troubleshooting through Human-AI Collaboration

TL;DR

3DPFIX tackles the challenge of remote novices troubleshooting 3D printing by transferring the rich, community-generated knowledge from online archives and forums into an AI-assisted troubleshooting workflow. It integrates image-based failure diagnosis, visual explanations via Grad-CAM, on-demand terminology, and a guided, two-tier solution flow driven by social annotation data. In a formative study, it identifies core design requirements, and in a summative study, demonstrates reduced cognitive load, improved failure identification, and enhanced knowledge acquisition compared to traditional online resources. The work advances human-AI collaboration in sensemaking tasks and offers a scalable dataset and design roadmap for future community-driven diagnostic tools in 3D printing and beyond.

Abstract

The widespread consumer-grade 3D printers and learning resources online enable novices to self-train in remote settings. While troubleshooting plays an essential part of 3D printing, the process remains challenging for many remote novices even with the help of well-developed online sources, such as online troubleshooting archives and online community help. We conducted a formative study with 76 active 3D printing users to learn how remote novices leverage online resources in troubleshooting and their challenges. We found that remote novices cannot fully utilize online resources. For example, the online archives statically provide general information, making it hard to search and relate their unique cases with existing descriptions. Online communities can potentially ease their struggles by providing more targeted suggestions, but a helper who can provide custom help is rather scarce, making it hard to obtain timely assistance. We propose 3DPFIX, an interactive 3D troubleshooting system powered by the pipeline to facilitate Human-AI Collaboration, designed to improve novices' 3D printing experiences and thus help them easily accumulate their domain knowledge. We built 3DPFIX that supports automated diagnosis and solution-seeking. 3DPFIX was built upon shared dialogues about failure cases from Q&A discourses accumulated in online communities. We leverage social annotations (i.e., comments) to build an annotated failure image dataset for AI classifiers and extract a solution pool. Our summative study revealed that using 3DPFIX helped participants spend significantly less effort in diagnosing failures and finding a more accurate solution than relying on their common practice. We also found that 3DPFIX users learn about 3D printing domain-specific knowledge. We discuss the implications of leveraging community-driven data in developing future Human-AI Collaboration designs.

Figures (12)

• Figure 1: (a) Yearly count of Q&A posts in FixMyPrint (2021 counts posts between Jan. and June), (b) Survey results of C3-3, users' changing perception regarding online communities depending on their self-rated expertise, (c) Survey results of C4-3, users' changing perception regarding online archives depending on their self-rated expertise
• Figure 2: This is the top section of the interface where users can select the 3D printing images they uploaded for diagnosis (upper subsection) and the corresponding failure type predictions with saliency maps generated by our models (lower subsection). Images are clickable tabs where red-colored borders indicate active selections. The system also displays the likelihood of each failure type prediction (Highly Likely: 75% - 100%, Likely: 50% - 75%, Unlikely: 25% - 50%, and Highly Unlikely: 0% - 25%). Users can click a specific failure prediction to explore further the solutions on the bottom section of the interface shown in Figure \ref{['figure:system_description']}, \ref{['figure:system_solutions']}, \ref{['figure:system_hoverover']}. Users can also use the toggle switch buttons to "Show All Failure Types" or "Show AI's Best Guess" which filters out the predictions below the 'Highly Likely' level.
• Figure 3: The bottom section of the interface has 3 tabs to investigate solutions for the selected failure type by the module in Figure \ref{['figure:system_diagnosis']}. The first tab, 'What's This Problem?' shows example photos about the common visual characteristics of a failure type, as well as an easy-understandable description on the right.
• Figure 4: Some technical terms in the solutions have detailed explanations and example photos when users hover over the terms (blue-colored with underlines). These underlined terms are also clickable on external websites with more comprehensive descriptions.
• Figure 5: The second and the third tabs of the bottom section are for investigating solutions. Users usually start with "See Common Solutions" where they can either search for existing clues in "Step 1. Find a clue." (on the left), or go straight to "Step 2. Learn solutions." (on the right) to browse the solution cards they think are relevant to their 3D printing failures. Solution cards are color-coded by their difficulty levels (Basic is green, Intermediate is in yellow, and Advanced is in red). A difficulty-level filter is also available. If users cannot find the solutions they need in the common solution tab, they can go to the last tab to see a more comprehensive batch of solutions provided by the system. All the navigation and filtering features are the same as in the previous tab.