Week 1

Week 2

Week 3

Week 4

Week 5

Week 7

Week 8

Week 9

Week 10

Week 6 (June 24 - 30)

For the prelim discussion group this week, I read the articles "Prototyping for Tiny Fingers" by Marc Rettig, "Direct Manipulation vs Interface Agents," excerpts of debates between Ben Shneiderman and Patti Maes, and "Principles of Mixed-Initiative User Interfaces" by Eric Horvitz.

"Prototyping for Tiny Fingers" was a quick read about utilizing lo-fi prototyping (simple cut-out paper prototypes) to get user software evaluations. Rettig argues that using lo-fi prototyping is quicker than creating the complete software prototype, it is also cheaper, most notably in man-hours, and it also allows more time to be spent on the ideas and concepts of the software, rather than the details of page layout, graphics, color, etc. Utilizing a lo-fi prototype before writing up the computer code often makes sense, and the discussion group agreed to this. Some members have used lo-fi prototypes in user evaluations and found the process quite easy. It of course itn't always appropriate, but it's an useful concept to keep in mind. We should consider doing this when when we want to work out design and content details of our nutrition device.

There was actually much consensus in the debate between Schneiderman (Direct Manipulation) and Pattie Maes (Interface Agents), because at the end both sides agreed that the direct manipulation and interface agents do not have to be mutually exclusive. Direct Manipulation has to do with allowing the user have ultimate control over all details of software procedures. Schneiderman stresses "information visualization" - the concept of allowing users the option of switching between seeing the big picture and details of data and procedures. Software agents traditionally have to do with having automated software procedures which act for the user. Maes argues that direct manipulation and software agents are not mutually exclusive largely because software agents are personalized to users' preferences and are not strictly autonomous agents.

The excerpts also included an interesting short discussion on the role of speech interfaces. Schneiderman notes that speech interfaces have important niches, most notably for use by disabled users and applications that require constant use of the hands and eyes. However, he points out that many speech proponents do not often recognize the scientific evidence that speaking is cognitively more demanding than pointing, in that is uses both short-term memory and working memory. He explains that you can do hand-eye tasks in parallel with problem solving more easily than you can speak while problem solving, and its an issue that the proponents need to work around. I had never heard of this before and found it very interesting, and would like to read more about it. Naes agreed with much of what Shneiderman said and concurred that speech is most useful in situations when the hands are already doing something else.

In "Principles of Mixed-Initiative User Interfaces" Eric Horvitz from Microsoft Research presented the LookOut program for scheduling and meeting management, a program which integrates both direct manipulation and interface agents. In a nutshell, the LookOut program is designed to scan e-mail and automatically insert meetings into Outlook. Horvitz goes into the details of how the program allows for users' direct manipulation but as the same time has many automated agent features. (I won't go into the details here) Most notably, in the discussion group, we talked about how the software was never evaluated, or at least Horvitz did not mention an user evaluation stage. We also talked about how the software was an interesting attempt to implement natural language processing.

I compiled the first two weeks of data this week and was a bit disappointed with the participant submission process. The participants were very late in submitting their second week of data. I stated on the schedule that they should mail the week's receipts and questionnaires within 1 to 2 days after the end of the week, but their envelopes were postmarked as being 4-5 days after the end of the week. Unfortunately, since we didn't state any repercussions for late submissions (we only stated that we wouldn't pay if they didn't submit anything), we still have to pay them the full amount for the week.

User Study Lesson 5: Consider awarding survey participants for timely submissions, and reducing payment for late submissions. It's frustrating when data is late, and money incentives may motivate users to submit data on time.

From the first week of data, the participants didn't buy many groceries and there was a poor correlation between what they bought and what they ate. They probably had groceries from the week before or consumed pantry items. The second week both participants bought more items, though it seemed like one participant may have had a dinner party/bbq, because a large amount of such foods were bought. We should ask participants if their receipts reflect buying food for guests.

From one of the participants receipts, I noticed that there is a major chain that uses some generic labels for food products (e.g."Grocery"). We don't want to include people in our survey who often purchase from this store, so I created a final questionnaire for candidates to break down by percentage where they usually shop. On this questionnaire I also asked the candidates to break down the number of times they eat out for lunch and dinner (this may give is a better estimate) and if they would be open to being taped during their interview.

It was a little difficult trying to figure out how I would store the receipt data and food consumption data. Since one of our primary goals is to look at the correlation between the two, I created an excel sheet (to be saved as CSV) based on the original food questionnaire. I entered the frequency of food consumption on one sheet and the purchase prices of foods on another sheet.