Skip to main content

Design I - 2026-02-26

Research answers questions but interventions insert something new into the world (they disrupt the world).

With interventions you do that STEPS IN RESEARCH diagram twice.

Design is the ability to imagine that-which-does-not-yet-exist, to make it appear in concrete form as a new, purposeful addition to the real world

— Nelson and Stolterman, The Design Way

Designers make choices. Design is a way of rearranging the world, you introduce objects, methods into the world that disrupt the world.

There’s a Basal Model of Design (Doblin, 1987): Some state —→ Transformation Function —→ Desired State.

Design is hard because you some kind of leap of faith: You have zero information on the future state. You are predicting it. This actually captures why Qualitiatve Research can only get you so far. You know what the world looks like, you still need to know how to change it.

John Chris Johns (1970) proposed three steps for that Transformation Function.

  1. Analysis: Understand current state of World. Break into pieces.
  2. Synthesis: Put pieces together in a different way
  3. Evaluation: Did it work?

Ogilvie and Liedtka (2011) have a different process:

  1. Discover: What is?
  2. Ideate: What if?
  3. Embodiment: What wows?
  4. Develop: What works?
  5. Evaluate: Did it work?

There is a negotation between discovering problems and filtering them through your own skillset so you can solve them appropriately.

There are times when pure qualitative research won’t do. You’ll need to engage in Contextual Design. It suggests five Work Models for any situation in any environment: Flow, Sequence, Artifact, Culture, Physical. Each model helps you find stuff to solve or work on.

  1. Flow: Communication and coordination between users necessary to do the work.
  2. Sequence: Break complex activities into sequences of steps. You may create some sort of consolidated sequence model after collecting sub-sequences.
  3. Artifact: What do you need to do your work?
  4. Culture: Discover culture, policies, values that help or hurt work.
  5. Physical: Actual structure of the space. Constraints. (You want to do a decision support system for rounds but they don’t have access to the system during rounds. There’s no point.)

Each one of these has various aspects to them. See lectures.

Discover​

Now there’s no real coding of data. Goal is not come out with findings but to come up with solutions. Now collect over many days, consolidate, and perform an Interpretation Session. Affinity Diagrams are great for this. Emerge clusters of things.

Ideate​

Generate a lot of ideas as possible. Make them different. “Kill your darlings”. Lots of ways to do this (IDEO Cards, Lateral thinking, Brainstorming, etc)

Toyota has multiple versions of every single automobile. They push it all the way to production so they don’t make premature decisions about what people like.

Embodiment​

There are a lot of ways in which you can express your design ideas (“Design Embodiments”). Some of them:

  • User stories/scenarios
  • Storyboards
  • Prototypes
  • Wizard of Oz1
  • Participatory Design

Note that the appropriate level of detail is important. Roughly speaking, each of these entail higher and higher levels of detail.

Mark Weiser

This is a very seminal HCI paper in a popular journal. Read it!

Design​

…

Evaluate​

Eval early and frequently. You can do this with varying levels:

  • Design Crits (just people on your team) - $0
  • Expert evaluation (hire them) - $$
  • User evaluation - $$$
  • Field evaluation - $$$$

There two biggies: Cognitive Walkthrough and Heuristic Evaluation.

Note that in heuristic evaluation you are asking people who are not experts in the domain to eval your design/system. Think of people who don’t know anything about kinesiology. You give them everything they know to complete the tasks and give them contexts and then have them eval.

“Recognition versus Recall”: Think of a GUI versus a CLI (you depend a lot on recall for the latter).

There’s a Nielsen Rating Scale. You don’t need more than about 6 evaluators (diminishing returns).

With Cognitive Walkthrough, just think of the ‘commonsense’ stuff you’d write down if you were asked to evaluate the UI/UX of an ATM (which all the young and cool kids use these days…)

Usability Evaluation/Testing​

You give people a system to use and interview them: this is not a Usability Evaluation. This is a user study.

This is a bit more formal/controlled. It happens in a semi-controlled settings. There are selected users, there’s scripts, they’re given detailed instructions, they’re observed, they’re questioned, their frustrations or joys are recorded. Here’s an NIST definition (note the number of times the word “specified” shows up):

The extent to which a product can be used by specified users to achieve specified goals with effectiveness, efficiency, and satisfaction in a specified context of use.

Feasibility Evaluation​

This is a limited/pilot deployment. Will users use your stuff? How will they react? Etc, etc.

Footnotes​

  1. This is a bit like a mechanical turk. User doesn’t see that a computer has actually done the work. Idea is that this was cheaper than software (lol). Danger here is that you can emulate very intelligent behaviour in your system that a computer might be incapable of. ↩