Jeremy Yuille has been working with the Australian Broadcasting Corporation (ABC) on a project to give viewers and listeners around the country a chance to upload their own content. Why would people upload their own content? The most cited reason for doing such a thing is, “it’s a place to display my work,” followed by “give my work a chance of being used by the ABC” and “get recognition from the ABC.” So Jeremy and his team (Chris Marmo, Reuben Stanton, Marius Foley) put the mental model together of various types of contributors, analyzed it in terms of how to support them, and created a site. Best of all, the team is sharing their work with the world.
Jeremy has generously shared the diagram with us here. (135Kb PDF).
Note how Jeremy, being a communications designer, reformatted the diagram to show mental spaces and towers as foundations to the boxes, rather than as headers. Jeremy blogged the details of this project here.
Team Twitter handles:
Jeremy Yuille @overlobe
Chris Marmo @kurisu
Reuben Stanton @absent
The Health 2.0 Conference is going on in San Francisco October 6-7 2009. IDEO is holding a contest to encourage and emphasize patient-centered design. Today I heard from Julie Cabinaw (@healthux, see her comment today on this article) that the first design her team at Healthwise created based on their mental models was accepted as one of 10 finalists in the contest. Julie says she hears more on Twitter about a competing product for creating prayer flags for child cancer patients, so don’t expect a win. But top 10 is a great place to be!
Congratulations, Julie and team!
Frequently when people hear the phrase “mental model” they think of a narrow slice of Don Norman’s definition. In his book The Design of Everyday Things, right after discussing his example of the refrigerator/freezer controls, Don Norman writes this definition of mental models: “the models people have of themselves, others, the environment, and the things with which they interact.” (page 17 in the First Doubleday/Currency Edition, 1990) Because many of us are interaction or information designers, we mostly focus on the “models people have of things with which they interact” part of the definition. It’s good to remember that Don’s definition is broader.
(Don Norman’s definition of a mental model.)
Mental model diagrams are your models of “others.” Not only that, these mental models are not necessarily just your model, but your whole design and development team’s model. I would like to define that term “others” more specifically, too. “Others” is a particular audience segment (or persona) trying to do a particular thing. For example, it is people who are “Passionate About a Topic” choosing which college to attend, or people who are “Exploring Paths” choosing a major to study. It is “evaluators” deciding which data management environment to invest in for their business. It is people who “Think It Through” considering how to meet people they might be interested in dating. It is people “Uncomfortable with The System” coping with a health problem.
(Your team’s mental model of others. You will want to be more specific than “customers.”)
Keep in mind you might want to model your internal employees and service providers as well. A hotel might want to model what front desk and concierge folks do. In addition to lunchtime patrons who want to “Shake It Up,” a fast food chain might study the employees making the burgers during the lunch rush. Realize you can model a wide range of groups. Your team can create several shared models that help you support and design for all sorts of various “others.”
And yes, the top half of these mental model diagrams is really the mental model part; the bottom half is where you align the support your organization offers people, whether they are internal or external. By matching services, processes, and information under each tower, you can see if they fit what people are trying to accomplish. You will see where there are strong and weak matches, or no matches at all.
When I first proposed the book title, the board of directors at Rosenfeld Media felt that a new term was needed. They suggested “alignment diagrams.” Early clients and readers convinced them otherwise.
Too much belief can be a bad thing … or at least misleading and lotsa-extra-work-inducing. Try not to confuse opinions and guesses with actual behavior-molding philosophies. As teams ramp up toward proficiency at combing items from the conversation transcripts, I’ve seen them comb something like eight mistaken beliefs for every one actual-real-live philosophy. Then the mistaken beliefs cause lots of over-thinking and gray hair during the grouping phase of the conversation analysis … and I really don’t want mental models to cause people gray hair.
Philosophies can really show us the motivation behind a behavior, providing reasons why someone does something. But philosophies are hard to pinpoint. My associate Eric Fain explained it this way to a client. “Because participants may not have verbalized a particular philosophy before being interviewed, they often make several attempts at explaining themselves. People speak colloquially and use phrases with ‘think’ or ‘believe’ in them as a way to soften an opinion or a simple explanation. Complaints can also be expressed similarly. For example, ‘I think our training program is not structured.'”
Take the time to think through what the user is saying. Try to avoid grabbing anything that has “believe” or “think” in it–more than likely the statement is an opinion, guess, conjecture, complaint, or statement of fact. Ask yourself the question, “Is this something the person thinks is going on, or is it something they know and hold as a guiding light in their life?”
I’m convinced that getting the labels “right” for the quotes will make the mental model method easier than many other analysis methods. The best labels keep you out of the traps that make your mind loop and loop over the same questions: “This could mean two things!” “This is factual, but what do I do with it?” “This needs to sound more academic, so my bosses will respect it.” As you are combing through the transcripts, you want to stay close to the participant talking, and you want to label all their quotes with a verb that is first-person, present tense.
A friend of mine, Karen Lindemann from Hamburg, Germany, is helping me out with a mental model about how people communicate with each other on a daily basis. When she read about verb+noun labels, she assumed I meant the infinitive form of the verb: “To tell,” “To decide,” “To worry.” It wasn’t working very well in her native language and felt clumsy to her, which is when I asked her to use first-person verbs and to bring the personal pronoun “I” into the label. “Now I think I understand why I had difficulties,” Karen laughs. Here are some of her new examples:
The most important thing to remember is to keep the verb first-person (“I”) present tense (“walk, sing, wonder, worry,” etc.). I frequently encounter people using third-person (“he, she”) and past tense (“walked, sang”) or gerund form (“wondering, worrying”). These examples put a barrier between you (the researcher) and the participant. Instead, I want you to step
through the barrier and become the participant for a while, when combing the transcripts.
The day after I get back from Barcelona, I’ll be going to UX Week 2009 in San Francisco. The four-day conference starts Tuesday September 15th and promises all sorts of good speakers. I don’t think I’ll make the opening speech due to jetlag (which means I will miss Aaron Forth’s speech about my beloved Mint, phooey), but I’ll try to be there later that first day to chat with folks. If all goes well, I’ll be there all day each of the following days of the conference. There are too many good speakers to miss, and I’m dying to tell Erin McKean I had glasses just like hers back in 1990!
(That is my best friend Carolyn Wan with me.)
If you’re going to be anywhere near SF that third week of September, you really ought to attend. And to sweeten the economic deal, I’ve got a discount to give you: enter INDI as a code when you register for 15% off. But-wait-there’s-more! If you’re a past or current client of mine, or related to a client of mine, or if we’ve worked together in any fashion, or if you’ve attended any of my workshops or presentations, contact me and I can get you an even better discount!
It is an unfortunate choice of words on my part, which may lead people astray during combing. The word “task” is leftover from the practice of “task analysis” which had been a favorite practice of usability research in the last decade. For the usability practice, a task is something someone does, like “Register for an account” or “Go to the doctor.” I explain in the book that I don’t really mean this kind of task. I want something like “Feel excited about the convenience of voice-controlled calendar access on my mobile phone via toktok” or “Ask the doctor if the online article about symptoms describe what I have.”
“So it’s not the act itself, but the underlying motivation,” surmises Voltaire Santos Miran, on of the founders of mStoner. He’s exactly right. And it includes emotions, reactions, and guiding philosophies. Voltaire goes on to say, “We’re thinking of calling it a ‘nugget’ rather than a task. A nugget is like mining for gold. Or, it’s like a little bite-sized chicken McNugget! Small and easy to eat.”
Combing is not analysis. Combing is the period of time when you must forget your role as a researcher or an employee and get to know each participant at a personal level, through their transcript or audio. At this point, you want to keep the label of a task (or “nugget”) as close as possible to the unique human nature of each participant. Represent their inner thoughts. Use their words. Try to capture what they were implying in the conversation. Save your higher-level labels for later. (Combing is also a great way to learn better interviewing skills, as you find parts of the transcript where you wish you could have more insight and ask deeper questions.)
Grouping is analysis. This is the point where you put your combed labels together and label the groups with higher-level phrases. When you have finished combing, then you can revert to your role as analyst and researcher.
If you haven’t started on your reading list yet, here’s a suggestion. Since the pile of books you want to go through is tall, and your time might be better spent with friends enjoying discussions, say, about the latest movies, try allowing yourself to read just two chapters a week–from separate books! From the Mental Models book, I recommend Chapter 8. This is the chapter that helps you see how different this type of analysis is. Instead of focusing on facts, you focus on motivators. What makes a person do something this way? Mental models are all about letting you see your customers from a whole new perspective, and this type of analysis is at the heart of it.
I’ve been guiding the fabulous folks at the University of Buffalo (and the team at their design partner mStoner) through the interviewing process this week. One of the university stakeholders for the project wanted to be interviewed as a participant–as someone who keeps track of what an organization is doing and crafts his decisions based on what he learns. As expected, the interview kept bouncing back to what this stakeholder does at the university, rather than branching out to similar habits he might have in other circumstances. When I tried to explore how he tracked information about other organizations than the university, he was surprised and said he wasn’t prepared to talk about other topics. It was disappointing because the other topic we glimpsed was unique to him, and I could sense that we would have been able to go deeper into what was motivating him.
That’s when it hit me. Perhaps instead of telling him we were going to interview him, we should have indicated it was going to be a more informal session. We should have said it would be conversational, where we talk about a variety of topics and follow branches in the stream of thought. Then the participant might have felt more relaxed about jumping to topics other than what he thought were “sanctioned” for the interview.
I mentioned to the team members that I thought the term “conversation” would be better to use than “interview.” They agreed and immediately changed the recruiting screener to use this word instead of “interview.” True, the definition of “conversation” is “an informal discussion or exchange of ideas,” which is not quite what we are doing, but it’s close. (We are listening rather than adding our ideas to the conversation.) The emphasis is on the informality of it. That’s what I like about it.
The Python script used to generate mental model diagrams has been updated to allow for added flexibility in output. Originally, box labels greater than about 48 characters would render outside the margins of the box, requiring close monitoring of the label length and often caused the label to be rewritten awkwardly or with abbreviations.
The new script was updated to allow for the following:
- Variable box height setting (on or off)
- Base box height setting
- Max box height setting
- Page aspect ratio may be altered, allowing the mental model diagram to be rendered in either landscape or portrait
- The boxes with the greatest height are moved to the top of the tower when rendered
- Tower height/width is rendered relative to the cumulative height of the boxes plus margins
Requirements for Running the Script
- Install Python version 2.6.2 for Windows (latest supported by PyWin)
- Install Pywin32 version pywin32-212.win32-py2.6.exe (Match the version number to the version number of Python you install–you will need to look at the list of exe files to find the right one, which may involve clicking on an upper-level folder; don’t just “download now” the whole package. Go for the matching exe file, with a strange file name like the one I show here.)
- Excel is required to run the script. The script does run with Excel 2003, but has not been tried with other versions
- While Omnigraffle or Visio are not required to run the script or create the diagram, they are required to view the diagram
- Note: As of Jan 2011, I have tried running Python27 on my Win7 32-bit OS, with the matching win32 Python extension, and it works. I have not tried Vista, egad.
Setting Variables Within the Script
Variables within the script may be set by the user to adjust the appearance of the mental model diagram. To set variables, open the script in any text editor such as NotePad or UltraEdit. There are numerous text editors available for free if you do not have one. Avoid rich text editors such as MS Word, or WordPad.
Caution: There are several variables that are calculated relative to other variables. Tower width is an example of this, where it uses the BOXWIDTH setting plus the spacing settings to render the Tower width. If you change these settings you are likely to break the script. If you do this, please download a fresh copy of this mental model script.
The first variable you may change is VariableHeight. Setting it to “False” restores the original script behavior, hiding all updates in this release. BOXHEIGHT defaults to 0.375 inches (3/8″). (You can reset this variable if you want.) At this default height, box labels will be restricted to about 48 characters and will overflow the box margins if too long. Setting VariableHeight to “True” allows the boxes to adjust automatically to the length of the content. (default = True)
The second variable you may change is PageLength. This variable allows the height of the diagram to be set in inches. Setting this variable to 11 for example will cause the diagram to render in portrait. Setting it to 8.5 will cause the diagram to render in landscape. (default = 11″ – portrait) One can assume that setting it to other numbers will allow rendering for slightly different page heights, although we have not tested it yet. Between this variable and MaxTowerHeight, one could possibly format the model to fit on plotter paper, admittedly still at the small font size.
Box Variables If VariableHeight=True:
- MAXBOXHEIGHT – Sets the maximum height in inches for the box height (default = 1.5 inches)
- MINBOXHEIGHT – Sets the minimum height in inches for the box height (default = 0.375 inches, 3/8″)
- BOXWIDTH – Sets the width of the Task boxes (default = 0.625 inches, 5/8″)
- Box2BoxSpacing – An empty space between boxes (default = 0.0625 inches, 1/16″)
Tower Height Variables If VariableHeight=True:
- MaxTowerHeight – Sets the maximum height for the tower (default = 4.05 inches)
- VariableTowerHeader – Turn on/off tower header content adjusting. Turned on (True) computes a space occupied by a tower header and adjust the boxes accordingly. Disabled (False) leaves a constant space controlled by a variable TowerHeaderExtra (default = True)
- MaxBoxNumberPerColumn – Sets the maximum number of boxes in a tower (default = 7)
- Box2TowerSpacing – A gap between a tower border and a box side (default = 0.125 inches, 1/8″)
- TowerHeaderExtra – Height of the tower header if VariableTowerHeader = False (default = 0.375 inches, 3/8″)
- If you are not using the Excel template, please be aware the the script is looking for column headings with the following labels in the exact order: Mental Space, Task Tower, Task
- Some characters will not render by this version of the script. Curly quotes will break the script, as well as some diacritical marks such as the acute acccent in fiancé.
- This script has been tested on the PC, but it has not been ported to the Mac.
Note: this list will be updated as new issues are found.