Design 10: Human-Computer Interaction

Theme

I first heard of Human-Computer Interaction at university when registering for elective computer science courses. I remember seeing posters about the course on bulletin boards in the engineering halls. I always thought it was an interesting subfield of CS because it seemed very product-oriented and visually creative. Human-Computer Interaction feels distinct because it's not technical in the sense of algorithms or theory. It's more about how people feel and interact with your software product, which felt novel when I first encountered it in undergrad.

I didn't end up taking the course, but I always thought it was really cool.

Grid

I didn't realize I messed up until I started grid construction. When creating puzzles, I write clues first because I value theme cohesion and clue clarity most. I felt really good after the clue creation process, but I didn't realize that all my clues were very long. For context, 11/14 clues were 7 letters or longer, with a median length of 8, and the grid is only 15x15. Since clues must be fully connected, this made fitting everything together more difficult.

A balanced grid didn't seem feasible, so I made it unbalanced, with most clues across and only a few were down. This resulted in 11 across clues and 3 down clues.

Visually, the grid looks "nice," but there aren't as many down intersections as I would have liked. I don't like heavily unbalanced grids since the lack of intersections makes it feel less like a crossword and more like a fill-in-the-blank. However, this was the best outcome I could find given the clues I selected.

Clues

I was originally worried about having too many people-centric clues because that makes puzzles very recall heavy. The clues include Shneiderman, Nielsen, Norman, and Card, with two other clues named after people, Hick's Law and Fitts' Law. Admittedly, 4 clues explicitly about people and 2 clues named after people is more than I would like. I think 3 people clues is a solid goal.

However, the 4 people I included felt foundational to HCI. Furthermore, even though the other 2 clues are named after people, they refer to widely known models in the field, so it didn't feel as heavy as just including more names.

For the remainder of the clues, I included important principles and concepts that are core to the field, like signifier, affordance, and visibility. I tried to make the questions explicit and highlight the key aspects of each concept in the context of HCI.

Tradeoffs

I didn't fully consider clue lengths when writing them, which ended up constraining the grid. As a result, I had to make a tradeoff between maintaining theme cohesion and achieving a well-functioning grid. I chose maintaining theme cohesion, resulting in a very unbalanced grid.

I also made a tradeoff in how I included Hick's Law and Fitts' Law. Originally, I wanted the answers to be Hick and Fitts, with the models referenced in the clues. However, that would make the puzzle too people-centric. I think Hickslaw and Fittslaw are slightly worse fills compared to Hick and Fitts, but since the models are so well known, I didn't think spelling them out explicitly would negatively affect the solving experience too much.

Notes

After publishing the puzzle early in the morning, I went back later that day to make minor refinements to the questions. Even though I spent a lot of time trying to make the clues clear and explicit before releasing, some still felt clunky when I reread them with fresh eyes. Refining the clues again took less than an hour but felt like a worthwhile improvement in clarity. It's difficult to write clear and specific clues in a single pass.