Nico Ritschel's Ph.D. Defense Summary

Nico Ritschel's research focuses on refining block-based programming by integrating elements from visual programming to make it more accessible and effective for end-users, especially in the robotics domain.

Problem Statement

• Block-based programming is mainly used for computer science education. Can they target other tasks, such as end-user programming?
• The challenge: end-users often need to write larger, real-world programs, contrasting with the simple toy examples students typically handle.
• Traditional block-based programming struggles with scalability, especially in terms of readability.
• While visual end-user programming tools like Excel and Simulink support bigger programs through domain-specific visual abstractions, creating new visual languages is difficult and costly.
• Solution Approach: Merge design features from visual programming into block-based programming languages.

Target Domain: Robotics

1. Current Scenario:
   • Professional tools exist, but they're challenging to use.
   • There needs to be more effective block-based tools in the domain.
2. Robot Arms for Factory Floors:
   • Task: Coordinate and synchronize two robot arms.
   • Issues: Current block-based languages require complex solutions like nontrivial mutexes.
   • Solution & Studies:
     • Proposed two design ideas:
       1. Represent programs for each arm vertically and side-by-side. Synchronized actions appear as shared nodes between the arms.
       2. A left-to-right flow resembling video editing.
     • The 'side-by-side' design was selected.
     • A study found that end-users using this design outperformed those using a commercial, text-based tool.
3. Mobile Robots for Warehouses & Labs:
   • Task: Handle large tasks across multiple workstations.
   • Issues:
     • Difficulty decomposing long programs and locating where to make changes.
   • Solutions & Features:
     • Introduced block-based language that supports functional decomposition.
     • Provided two separate canvases: one for task composition/movement and the other for low-level task definitions.
     • Included triggers as dataflow graphs to improve the visibility of nested expressions and enhance user freedom in structuring programs.

Questions Addressed During the Practice Session

1. Why focus on the two robotics scenarios?
   • They are important and relevant in the robotics domain.
   • These scenarios present challenges for end-users learning to program.
   • They represent a complex form of programming that's worth refining.
2. Would functional programming principles enhance end-user visual programming, given the imperative nature of block-based programming?
   • The inherent complexity in robotics means many elements can't be simplified.
   • Introducing functional programming might not necessarily boost user productivity.
3. What was the environment for user studies?
   • Engaged actual end-users for genuine feedback.
   • Also recruited students from non-computer science departments for a broader perspective.

Questions Asked During the Ph.D. Defense

• How were the visions and observations formulated?
  • Separate users into traditional versus new environments and then compare.
  • Gain knowledge of their needs and patterns.
  • Test on a small pool of users to refine the design.
• How do you account for the spectrum of end-users regarding programming experience, domain-specific task time, and tool experience?
• Which results were the most and least robust?
• What factors made the tool easy to learn?
  • The "blocks" concept is already well-known.
  • The tool matches the users' previous domain-specific knowledge (e.g., separate columns for two arms).
• How realistic is the decomposition at scale? Any evidence from related work?
  • More of a "lower bound," limited by the time of the user study.
• Why was the comparison made between block-based methods and graph-based methods?
  • Graph-based methods are already used in end-user programming, such as game programming.
• What is the importance and implication of the determined p-value?
  • We have a null hypothesis - there is no difference between the performance of the two groups.
• What improvements (e.g., 5%) are worthwhile?
• What are the advantages of block-based approaches over dataflow, and how can this be further investigated?
  • Different aspects, e.g., reading vs writing
  • Different domains, e.g., robotics vs game
  • Different styles of programs
  • Different representations of graphs
• How are potential accessibility challenges addressed?
  • Already addressed to a degree in the normal block-based domain.
  • Domain-specific challenges are directions for future work.
• How does the new tool compare with LLMs?
  • Can work together.
  • Have advantages in evolution and understanding vs. writing something that would work the first time.
    • Debugging.
    • Reliability.
    • No training required.
• What follow-up studies are anticipated for real-world usage? How do you anticipate the tool's usability in practical scenarios? Follow-up studies based on real-world usage in the wild may encounter unanticipated, really specific problems. Is your tool something someone wants to use in practice?
• Would featuring a table of reactive values a la Excel be beneficial?
• How do different domains within computer science influence the tool's design and analysis? What interdisciplinary expertise would be beneficial?
  • Information visualization.
  • Designing design drafts with an expert in visualization would be beneficial.
• What about your tool's applicability to expert programmers instead of end users?
  • Different design goals.