All Design is Redesign

I'm exploring website redesign, specifically with an eye towards end-user redesign. With increasing hordes of diverse people, devices and activities having virtual presence, leveraging motivated/power users to redesign existing systems for specific devices/uses/communities/etc may improve the interoperability between them.

Many proponents of web services suggest that 3rd parties will fill various niches if/when given programmatic access to an existing system. My instinct is in a similar vein, but requiring something less than "software engineer" level skills to play.

Here is an example of the bay area transit trip planner ( http://transit.511.org/tripplanner/index.asp ) redesigned for small-scale devices, like mobile phones -> http://liquidcycles.com/trip

Try it out -- what works/doesn't?

(thanks to Larry Leifer for the mantra-title)

Explanation (now with Spiritual Awakening!)

This blog was originally a replacement for the "idea log" requirement of cs247 ("interaction design studio") when I broke both my hands getting run over by a truck and had trouble writing. It will now rise from the ashes as a blog-phoenix of thoughts, insights and ramblings about interaction design and its siblings.

If you're reading, have fun -- and thanks!

CS247 Reflection

Has it only been 10 weeks? It seems like a lot longer -- in a good way :)

One thing that was stark was the dramatic increased in complexity when designing with more people. Our team went from two to five and it became nearly impossible to get everyone together -- making decisions slower and less definitive, lowering our efficiency, and making the project much more static and less experimental. On the other hand, it was great to be able to really drill down on a specific area and not worry that nothing else was getting down (I spent quite a bit of time learning about power-transistors and how to control higher-power devices with low-power control voltages)

From an engineering perspective, I became tangibly familiar with how difficult it is for programs to definitively get information about the "outside world". Using the desktop is like playing in the computer's sandbox -- it controls everything and you just get to play. Outside the desktop, the world is much more complex and subtle and even if the computer can get data -- its not clear what it means exactly.

Thanks for a great course!

Passively Powered Sensors

Owing to my obsession with cable tangles (here, for example, or here) and specifically my experience with the squeezers, I'd love to see if passively powered sensors could be developed (using electric induction, like RFID tags, or mechanical energy, like old wrist-watches).

Beyond cables, this could also break down some of the power hurdles of ubiquitous computing.

Photos from the Final Project

Click here -- don't miss Donald Norman at the Squeeze booth :)

Users as Generators

Our user-testing was excellent (not that they didn't have complaints -- but we got a ton of useful feedback) and, beyond this, they were eager to explore the re-purposing of the squeeze-interface for their own needs.

Obviously this didn't help us accomplish the goals of the IDM project, but it was highly useful, and inspiring, for us as designers/builders.

For example, one of our teachers thought the squeezer would be great for "English Language Learners" -- that is, kids who do not speak english well. Apparently, they often are shy about raising their hand because language is such a barrier to comfort (more than history, science etc) and a non-interruptive way to contact the teacher might be a promising addition.

I'm not sure if this can be expected in every set of user-subjects -- as teachers are often unusually creative, and our testers are all excellent teachers -- but it was an great experience in this case.

Sleeping Cars

Many cars make a mesmerizing sleepy sound (almost like soft snoring -- white-noise like, but with some rhythm) that is self-similar with the zombie-like state that their driver's often inhabit. Further, in discussions with others, I have found that successful commuters (ones that don't grow to hate it) do, in fact, enter a restful and relaxing state where their mind is slowed down and not particularly active.

I don't currently have an immediate design related to this observation, but will enter a restful zombie-like state until I do :)

Squeezable Cable Tangles!

Inspired by the evergrowing tangles of cables everywhere, I think it would be cool to make a squeezer out of a big squishy tangle of cables.

While this doesn't accomplish the needs of the IDM (Improved Decision Making) project, it might be very satisfying for potential first-world squeezers to exact some revenge on their evergrowing tangles of video/audio/computer/etc cables.. while, at the same time, directing a system according to their needs.

Emotional Mirrors?

Inspired by the magic-mirror project by Dan, Noah and Sim, and their difficulties building an interface that users can intentionally use without getting false positives (changing the display by combing your hair, etc) -- I wonder if a magic mirror could be a more subtle interaction.. it could combine elements such as the beat/volume/etc of your music, the color of your clothing, tone of voice, posture, etc to display an ambient colored light.. an "emotional mirror"?

The interaction would be really subtle, but potentially engaging.

This idea is also inspired by last quarter's HCI seminar in "Reality Mining"

Visualizing Aggregate Squeeze

For the IDM (Improved Decision Making) project, one of the problems of collecting squeeze data is displaying it with the following constraints:
1) The teacher may not want to be at the computer to know the current squeeze-interest
2) It may not be appropriate to display each individual users squeeze, but it would be good to get more than a one-dimensional value for the aggregate squeeze interest.

Above is an exhibit from the SF MOMA that uses a fan, fabric and weights to create a dynamic display. Using a similar system, but with multiple fans, we might be able to display more than one-dimensional squeeze information without losing any privacy.

On the other hand, it might be really confusing :)

P4 Ideation

Our P4 process included iterations in two main categories (with only minor software changes) ->

1. The Squeezers themselves. We experimented with a wide-range of squeezables: pieces of cardboard, raisin boxes, water bottles, staplers, sponges, pliers, tennis balls, bubble wrap, tangles of cable, etc. We selected our favorites among these items and more and tested them with Stanford students. The favorite was a specific pair of pliers (pictured above) and we used this as our final squeezable (we also instrumented a little-stapler squeezer and it was less-preferred, without question, to the pliers, by our test users)
2. The aggregate visualization. We discussed the visualization quite a bit, for example, this post describes one idea with fans and fabric and this . One problem that we encountered was the need for more-than-usb power-levels. While we did eventually get this resolved, it was only at the very end of the project, so we used a simple light-based visualizer for our user-testing. The 2nd image above shows both the light-based visualizer and the fan blown pinwheel visualizer.
   

LCD Polarization

Laptops with LCD screens are, for some reason, polarized vertically (only vertical light waves are emitted). Unfortunately, polarized sun glasses are polarized to block light in this direction because it cuts down on most glare from the sun (any light that reflects off a horizontal surface, I believe). Essentially, LCD screens compete in the same dimension as reflected sunlight.

If laptop LCD screens emitted light polarized in the other direction, a user could wear polarized sun-glasses and see a relative brightening of the screen compared to the ambient reflected sunlight (aka "glare")

Flickr Fumble

While I like the flickr interface on the whole, its response to multiple tags entered into the search-field is non-standard, un-useful and annoying.

For example, in the first screenshot, I search for "van roadtrip" looking for images that are tagged with both "van" and "roadtrip". But, in the second shot, you can see the flickr concatenates these terms and fails to find any photos with the tag "vanroadtrip"

Personally, I don't see how this is ever useful (it can be avoided by going to the "advanced search" screen and choosing "all of these tags" but I think that should be the default behavior (like web search engines)

Adding Search Terms to the Menubar of Web Browsers

Web browsers typically show the title of the web page the user is viewing in the menubar of the application.

However, often, the user reached the page using a specific search term and the search term is a better reminder of what the user was doing than the title of the page selected by the author.

For example, in the image above, I search for "'computer vision' emotion" but the title is "Vision and Modelling Group Research Projects" which is much less descriptive and less personal to me (because I didn't choose it)

This change/addition may be especially relevant when a user is switching back and forth between different web-browser windows because, typically, the menubar text is displayed by the window-manager/operating-system and the user is trying to quickly identify which browser window is which.

This idea was generated by my own personal experience using multiple web-browser windows to find information in an interleaved manner.

Followup on Social Buildings

These pictures from the Aperature Project are an excellent example of what I referred to in my Social Buildings idea.

For most buildings, I think the outside representation would have to be more abstract/ambiguous for privacy reasons -- but maybe also more interesting and intriguing?

Following Bill Gates

While discussing Computer Vision applications with a friend, I mentioned my simple idea for monitoring parking lots. Apparently, Bill Gates and Paul Allen got started with Traf-O-Data -- a system to count cars passing a point on the road.

Ridiculous!

Clustering Algorithms for Quick and Dirty Market Segment Analysis

Given how feature-driven many consumer products are (phones, cameras, cars, laptops, etc), it would be relatively easy to "select features" and do cluster-analysis for a quantitative look at offerings for a current market segment.

There is a lot of work on using clustering algos on customers ("conjoint analysis", etc), but a clustering of existing products by feature could yield a much quicker and cheaper analysis of existing market segments than extensive user analysis (somewhat analogously to the relationship between heuristic evaluation and actual user-testing)

When Human Computer Interaction is not about the GUI

Many of the biggest advancements achieved by web applications are not achieved by the graphical user interface (typically web-apps are more limited than native applications in this area), but rather the shift from a single-user interaction (like Word, Excel, etc) to multi-user interaction (ebay, paypal, friendster, flickr, etc) -- that is, the human-computer interaction is primarily human-human interaction, mediated by the computer.

Ebay, for example, still has a very limited graphical user interface -- essentially text with pictures -- but is arguably the most successful web application ever. Similarly, Craigslist is nearly a terminal-based application (text-only), but is immensely useful and popular because of its ability to mediate interaction between people.

Since these applications are special for their multi-user interactions not their graphical user interfaces, a good web-application prototyping tool would focus on demonstrating multi-user interactions more than building a good GUI.

Regarding specifics, I think this problem may be similar to the one schema building/analyzing tools face, where, the problem is primarily defining relationships between data elements. In this way, starting with a SQL schema, an additional set of rendering hints could be maintained from which a multi-user web-app prototype could be generated and tested very quickly.

This idea was an offshoot of my Flash improvement ideas.

Computer Vision in Parking Lots

Many parking lots have to hire people to track payment (or make sure that parkers are customers). These sorts of problems are a ripe area for computer vision algorithms because, while a computer may not be able to tell you what type of car is there, it can certainly tell you if there is a car in a given spot or not. And, further aiding adoption, in many parking lots there are already surveillance cameras installed.

For example, if a car is parked for 10 hours in a large store's parking lot -- the program could highlight it for a once-a-day check. Currently, there is no good way to get this information without continuous human monitoring of a parking lot.

Similarly, in other situations, the same information could be used to assign each new car a slot.. rather than having each car drive around looking for a spot (parking garages can get very jammed up)

Power of the Network.. err, Tangle

Pictured above is Scott Doorley trying to hold up a tangle of power-supply cords at the Computer Recycling Center -- you can see him leaning back to counter-balance the weight of the tangle. After tugging and untangling for five minutes, we eventually gave up on the power supply we were trying to reach.

While this is an extreme case, the proliferation of computer-connected devices/sensors/actuators/etc means that each person's individual cable-tangle grows over time.

Is there anything that can be done to address this growing problem?

Bee-Flower Structures at the Farmer's Market

Much like a flower that needs to both lure in a bee and keep it stuck long enough to brush pollen on and off, the farmer's market needs to lure in customers and keep them long enough to buy produce.

On the market-scale, this structure is evident in the layout of the individual stands in a looping structure with only a couple "exits". On the invidual stand-scale, this is evident from the U-style table structure (pictured above) where there are two ways into the stand, but once you're in, its not so easy to just walk out -- especially if there are more customers coming in behind you.

The Coffee Mug Tree

Following the need of garbage improvement, and inspired by the bee-flower structure of the farmer's market, our group came up with a coffee mug tree that customers could put their reusable coffee mugs on when they are done. This action returns their one-dollar deposit and may invoke the "tree falling" mechanism where all the mugs get dumped off the tree and into the mug-washer (further attracting the bee/customers to the flower/farmers-market)

For customers who don't return their mugs, they will be advertising the catchy farmer's market logo wherever they use it in the future :)

Garbage at the Farmer's Market

There is a lot of garbage at the farmer's market. The top picture shows some of the vendor-garbage and the one below it shows the large number of customer coffee cups. It was a cold day and nearly everyone had coffee, but the waste is not congruent with the "sustainable" spirit of the farmer's market (as explicitly mentioned by the sponsors and, otherwise, implicitly evident in the market itself)

Is there a way to improve the garbage situation using more reusable or, at least, recyclable items?

Farmer's Market Picture Diary

CS 247 went to the Farmer's market on January 21st, 2006. Our group went early and watched the market setup and get started -- my pictures!

Personal Marketing for a Personal Experience

The farmer's market has a highly personal aspect that would be difficult to maintain with traditional marketing mechanisms.

But, personal notes and recommendations from the farmer's market, floated to lucky recipients by helium balloons, might preserve and propagate the personal nature of the farmer's market experience while providing fun for both the creator and finder.

The littering aspect of helium balloons may prevent this specific idea from being popular, but apparently, latex balloons are both organic and biodegrable!

Social Buildings

Many buildings emit a cold exterior even though the people inside them are warm, and ironically, often feeling cut off from the rest of the world that feels cut off from them. It would be interesting trying to break down the building-wall by offering outsiders a feel for the insiders, without infringing on the insiders privacy.. and giving insiders a view of outsiders without being overly distracting.

Motion and/or infrared sensors could be used to detect movement both inside and outside and neon lights (or other lights/sounds/wind/etc??) could provide an abstract representation on the outside -- maybe with cartoonlike outlines or large pixelated reverberation. Similarly, pinwheels (credit to Hiroshi Ishii) or other ambient actuators could allow people on the inside to "feel" motion on the outside.

Squeezer as a Shared Artifact from the Policymakers to the Children

Similar to Deepti's idea of using the squeezer as a shared artifact from the children to the policymakers it might be possible to do it the other way too.

For example, footage of a famous policymaker (Al Gore or a philanthropist sports figure?) using the squeezer to watch a video clip could be used as an introduction to the squeezer for the children. This would both, entice them to use the squeezer, and highlight the importance of their task.

Along the same lines, it seems like uniquely and obviously identifying each squeezer might be a good idea. It could help highlight the unique role of each participant.

Using the Squeezer to Measure Policymakers Interest in Children's Documentaries

Similar in spirit to using common reference point to help share values, Deepti suggested that some policymakers might be willing to use the squeezer to record their reaction to the various documentaries. In addition to being valuable feedback, it might allow a more tangible connection between the policymaker and the children who made the film -- through a common squeezing device.

More than anything, I think this might have some ramifications for the aesthetics of the squeezing device.

Squeezable Film Editing

Based on previous IDM (Improved Decision Making) ideas of tagging-video and tangible-video-editing it might be possible to allow squeezable film tagging. This would involve users squeezing a squeezer to indicate their relative interest at any given time. For the IDM project, this would allow three advantages of:

1) Allowing the children to save video clips in a very time-conservative manner (the whole project is 4-5 days) -- and help keep them more engaged in watching the videos. Also, this allows preliminary editing to occur without a computer for each student.
2) Allowing the teacher to measure ongoing interest in a very non-obtrusive way and ask a specific student if they squeeze hard during some scene or be aware of sleeping non-squeezers.

Aggregate Process Artifact from the Squeezable UI Prototype for Film Learning/Making

Here are some of the highlights of our process that I haven't put in separate entries.

1) A crucial aspect of the Squeeze film editing model were the relationships teacher and student views. In order to capture this aspect in a single prototype, we built a meta-interface to allow users to switch back and forth between the teacher and student view.

2) Our Heuristic Evaluation revealed the following
i) There was a lot of confusion about the meta-interface. We modified this a bit visually and also focused more on explaining it to subsequent users. Because the meta-interface isn't testing the actual interaction we're interested in, we tried to keep it simple, but discrete from the rest of the prototype.
ii) The video playhead should be movable, aka allow scrubbing. We intended this, but its hard in flash.
iii) The "+{" aggregate icon was confusing. We shaded it darker for later testing and didn't get any more complaints/confusion (the aggregate display seems to explain itself).
iv) There needs to be squeeze-feedback in Student-mode. This was one of the stronger responses and we added it to the meta-interface for the final prototype.

3) In testing with Deepti, she suggested that the policymakers might be willing to use the squeezer too (she even thought it might even help enrich the experience for them). This led to the idea that the policymakers could even, through recorded video, teach the students how to use the squeezer. Both of these represent substantial revisions to our previous storyboard, which, in itself, represented multiple rounds of changes since our initial storyboard of tangible video editing.

4) Personally, I am very happy with our work during this stage. Among other things, I have seen that:
i) Scott and I work very well together
ii) Flash, despite its reputation, is not great for prototypes. At a certain complexity, it inhibits changes (b/c its so easy to break unrelated pieces, and so difficult to debug them) and therefore inhibits actual iterative prototyping.
iii) Given the nature of the project, its very difficult to actually do user-centered design. In our case, since the more different the children are from the western 1st world, arguably, the more important the documentary (since the policymakers will have less idea what life is like) -- we focused on the lowest common denominator.. a squeezer and a movie. Its been an interesting exercise, but in some ways it is only user-centered in one direction (we avoided dependencies, but did not leverage opportunities)

Squeezable Cameras?

With the rapid proliferation of squeezable interfaces.... :)

Both video and audio can be recorded very easily using a one-control interface.. is it possible that people might want the tactile thrill of squeezing their camera?

Hard to say.. but I'd be interested in trying it. There is a lot of video/audio functionality in newer phones, but it can be really cumbersome to use because its all packed into such a small device (even for young technophiles with good dexterity). Maybe there is another segment of the population that would love to just record-by-squeezing and then plug the recorder into their computer to hear/watch their recordings on a larger-format device later (esp with storage space an ever-diminishing concern)

Yogaballs as Mice

Using a mouse, even a nice one, is a limited-range/limited-muscle tedious effort. Manipulating the cursor by leaning different directions (or bouncing to double-click!) on your yogaball/seat might be a welcome break and a much more full-body experience... While it probably wouldn't be used all the time -- it could allow people to get into the flow more and maybe even think more physically when they're using the computer when they choose to use it.

Also, it would allow the user to leave both hands on the keyboard while moving the cursor -- mouse-keyboard commuting is a very common slowdown and physically aggravating factor.

Technical considerations:

From a functionality standpoint, the "yoga ball pointing device" would be most similar to the "pointing sticks" found on some laptops... where the computer moves the cursor according to the force, not location (eg: leaning farther over, moves the cursor faster)

There would probably have to be a way to engage/disengage the yoga ball -- otherwise inadvertant movements would move or "click" the cursor.

The Designer's Paradox

In a designer's utopia, we're completely freed from thinking about our tools because they are so well designed. But then, what ARE we supposed to think about? Designer's don't know -- thats for sure!

The point is simply that as a designer, you have to think about tools a lot so that others don't have to. But, in a twisted way, a designer hides the beautiful complexity of options that they themselves enjoy thoroughly.

Relatedly, a big part of a person's (even a non-designer's) pleasure in life might be the tools they choose and how they use them. Manual stick shifts are a good example of this. For these cases, the standards of design may need to be adjusted towards the playful (or other goal) rather than the simple.

Improvements on Macromedia Flash?

Existing problems with Flash include the following:

1) Extremely difficult to debug
2) Relatively high demands on end-users (have to install correct plugin/version)
3) GUI is extremely confusing
4) There are a huge number of entities and they are highly non-intuitive
5) Global namespace severely limits the size of potential projects and parallel development

All this combines to make a tool that has a fairly high threshold and an extremely low ceiling and that really isn't good for prototype-development at all. The whole point of developing prototypes is that you can change things quickly and easily -- and this is simply impossible for Flash projects of any reasonable size.

On the other hand, Flash is a "clip oriented" language and, in this sense, is good for developing small highly visual applications. An improvement would have to build on this strong point.

A improved system could:
1) Retain a clip-oriented aspect (vague -- needs more consideration)
2) Put all types under a single consistent object hierarchy
3) support non-global variables for larger/more-parallel projects
4) Allow optional stricter compling (like "use strict" in perl)
5) Output to javascript instead of .swf, allowing for easier access by end-users and easier/better integration with other website-components
6) A simplified GUI (vague -- needs more consideration)

Picture-Oriented Camera

Almost all digital cameras are focused on giving the user the most technical camera features like shutter/aperature/iso/flash and/or a large set of automatic shooting modes.

However, many people are overwhelmed by their cameras and don't know what the features are, much less how to use them. Basically, I propose, they are more interested in taking good pictures than learning about their feature-laden cameras.

Since the technical details of photography are only part of the equation, and auto-modes are fairly well advanced, I believe a picture-oriented (rather than featured-oriented) camera might be useful and more relevant than the evergrowing stack of technical features. It would "look" at the digital image and make suggestions and potentially highlight features in the frame to make the suggestions more clear.

A picture-oriented camera might make suggestions like:
1) The light source is behind the highlighted subject, you might consider moving it unless you want a silhouette.
2) The subject's (outlined) are small in the frame, you might move or zoom closer (this is one of the most common photographic mistakes)
3) The bright spot (outlined) is going to "burn" the photo (or a dark spot is going to blacken), you might move the frame.
4) You need more light to avoid blurring (there could even be sensor to measure the camera shake)
5) If you can rest the camera on something, you can lessen the blur.
6) Consider dividing up the key elements of the frame onto the highlighted marks ("rule of thirds")
7) Your flash will not reach the subject (outlined) consider moving closer.

It would probably be necessary to adjust the "comment level" of the camera (including turning off all comments). At a minimum, comments could be broken down into Lighting and Composition -- both of which are frequent components of bad photographs. And maybe more?

Similarly, a "suggester" would have to be closely calibrated with the limits of the camera. But, if successful, the results could be dramatic.

A big part of the challenge is to make the suggestions simple and useful, but not distracting or annoying. I'm going to consider this aspect more.

Progress on the Travelling Yuppie Problem?

With the ubiquity of infinitely growing TODO lists among busy professionals, we might be able to borrow from the field of action-planning to opportunistically allow people to be more efficient.

For example, if a person had the following items on their TODO list:
1) Return curtains to Home Depot
2) Drop off Jill and Bobby's wedding photos
3) Stop by Maz Auto to figure out why my windshield squeaks
4) Return Visual Thinking to Green Library by 3/1/2006
5) Check out the Wangechi Mutu exhibit at SFMOMA b/w 12/16/2005 and 4/2/2006

Then, the following sorts of queries could be run ->
1) I'll be free in San Francisco from 3 - 5pm today -- is there anything you (TODO) suggest?
A:
i) The SFMoma is open and you wanted to see the Wangechi Mutu exhibit
ii) Bobby and Jill live in SF and you need to drop off their wedding photos.
iii) There is a Maz auto at 3rd and Townsend, you wanted to get your windshield checked out.

2) I'm driving from San Jose to Stanford today, is there anything I should remember?
i) You need to return Visual Thinking to Green Library before 3/1/2006
ii) There is a Home Depot right off the 101, if you want to return your curtains

There are a lot of other kinds of queries that could be executed with similar knowledge (so-and-so is going on vacation, is there anything I need to accomplish with them? i need to do the following N things, what is the fastest way to accomplish them?)

But, minimally, as proposed, the system would require knowledge about:
1) time
2) space
3) people
4) commercial entities (store chains, etc)

Picturecodes

Idea: Using images that are both human and machine readable, physical links (Picturecodes) could be created from the "real" world into its virtual counterpart without being limited by the aesthetics of urls and barcodes.

Inspiration: The cooperative connection DV project is trying to connect buyers of skirts in a more intimate way with their Indian producers. One way might be to have a Picturecode of the producer or the village, etc that the consumer could scan using a camera(phone) and get linked into a producer's website.

Offshoots:

1. Uniqueness: The tags could be unique so that from the producer standpoint when a consumer scans in a tag, this information could be used for registration (avoid counterfeiting, customer preferences, etc)
   
2. Feedback: Once scanned, a consumer could provide feedback about the individual item. Because the items are hand-made they might even be able to make a custom order for a new item at minimally more cost (thus providing the consumer with a high level of service, the producer with "producer loyalty" and both with a stronger producer-consumer connection). Also, this would provide the producers with feedback in order to respond to the skirt market.
   
3. Motivation: As certain producers gain reputations for exceptional quality skirts, producers will be motivated to raise their quality. In addition to providing producers with a higher level of self-determination, it will make the cooperative connection skirts more competitive in the marketplace (where they are charging a premium price)
4. Aesthetic display: In the consumer store the Picturetag could be used for all sorts of interesting purposes -- for example, highlighting a link on a big "earth screen" between the producers location and the consumer's (maybe the link remains after the purchase, but fades over time)
   

Problems: Machine vision is tricky. It would be hard to use an imperceptible watermark.. more likely is a distinctive border or something like that.

Coffeehouse + Mass Transit - Driving

Very few things are more ubiquitous in the American morning routine than drinking coffee and driving to work. Maybe they could be combined and improved into a mobile coffeehouse?

People spend a lot of time and money on cars and coffee. The current reimbursement rate for gasoline is 48.5 cents, even on a commute as short as 10 miles + a cup of coffee ~= 10$

Add in wifi access and the ability to meet with people, and the proposition might be attractive.

Non-technically, I think a key aspect would be to recreate the warm ambience of a coffeeshop on a bus or train car. And, it is important to pick good routes.

Technically, a main challenge would be to have automated coffee generation in such a way that it doesn't offend the coffeehouse spirit (maybe a transparent coffee system that runs throughout the bus?). Also, automated payment and spill-proof cups would be necessary.