October 2009

only in an experimental psychology paper can you find...

"The media were demonstrated by having the participants solve three simple communication tasks by means of video telephone, hands-free telephone, and text chat. Face-to-face (F-t-f) was assumed familiar."

Schliemann, T., Asting, T., Følstad, A., & Heim, J. (2002). Medium preference and medium effects in person-person communication. In CHI '02 extended abstracts on Human factors in computing systems (pp. 710-711). Minneapolis, Minnesota, USA: ACM. doi: 10.1145/506443.506559

design research 101: you don't have to look at all the data. given a compromise between examining and categorising every transcript, photo, answer or looking at the data in different ways, pick manipulation & representation over organization every time.

Having lots of data points doesn't mean you have greater understanding: it simply means you have lots of data points and more chances to prove a statement true or false. Having fully organized data is great if you want to test conclusions, but it's easier to check hypotheses than to find a new, more powerful, more insightful way of looking at the phenomenon being studied. Finding a good explanation for behavior usually involves multiple rounds of trying to make sense of something. That takes time.

If you're short on time, forget trying to organize everything and spend more time interpreting and deriving conclusions from the data. After all, that's what you were hired for, isn't it?

Today I came home and noticed this package lying on the porch (I vaguely remembered having seen it for a few days now). Turned out it was a promotional campaign for Dunkin' Donuts, in aid of cancer research. And, oh, we'll sell you a free bagel.

And a free drink if you come in. In this nice sturdy plastic cup.

Only problem is, I didn't ask for a coffee cup!

So, in order for the Stefanie Spielman Cancer Research Fund to get $1, I have to buy 13 bagels from Dunkin' Donuts. No wait, it gets worse.

In order for me to buy 13 bagels so that the Stefanie Spielman Cancer Research Fund gets $1, Dunkin' Donut:
# pays for 1 coffee
# pays for the production of one coffee cup (for which there is no market demand)
# pays for the design & printing of a magazine-quality information sheet
# pays for the (non-reusable) plastic bag it all came in
# pays for people to bring this to my doorstep

In addition, I am forced to deal with junk I didn't ask for, and the planet is awarded one ugly and unwanted coffee cup that is too hard to destroy. Now I hate waste, so I'll try to recycle it. Which means that Dunkin' Donuts is forcing me to pay some of the costs of their marketing campaign. (And you, and your children, too - plastic takes forever to degrade).

By now I'm finding it hard to believe that the true unit cost of all this is less than a dollar, in which case more - probably much more - than a dollar was spent to increase the coffers of the cancer research fund by $1, and the coffers of Dunkin' Donuts by the profit from the sale of 13 bagels.

Outrageous. Things that Dunkin' Donuts could have done instead of this horror:
1. Advertised the promotion in-store
2. Offered a free coffee if you brought your own cup
3. Just given the Cancer Research Fund $1 multiplied by the campaign's estimated rate of conversion. And then have advertised the good deed.
4. Advertised it online (update: they did)

Instead, this comes off as a desperate attempt to get me to give my money to Dunkin' Donuts instead of a kind, charitable gesture. And one that does more harm to everyone than good.

Unfortunately, I don't know how to put dollar-values to each of the above costs. If you know how to investigate this further, please drop me a line. Also, there must have been some set of conditions under which this sounds like a good idea: if anyone can throw some light, I'd be much obliged.

a post on presenting with an active backchannel: but what if you were twittering during your own presentation? you could create a depth of interactivity that wouldn't otherwise exist... participating in the backchannel to unpack your ideas, reveal sources, create connections, explain uncommon concepts. twitter plugin for powerpoint: where are you?

PS: this is why we need activity-based computing and a more nuanced awareness of context.

a newspaper dispenser for the Columbus Dispatch, outside a Subway a newspaper dispenser for the Columbus Dispatch, outside a Subway

Here's an example of one my favourite ethnographic analytic moves: 'unpacking'. Unpacking is one kind of interpretation, where an observation is examined for unstated (or implicit) assumptions, consequences, and meanings. Here's how it works.

Consider the above image: a newspaper dispenser outside a subway. Here's what we're going to assume about it:

  1. It is intended to sell newspapers
  2. It is designed to support this intention

Now, notice the statement on the bottom of the dispenser: "We capture history every day". Unpacking this involves finding premises that give this statement meaning:

  1. "History" happens every day
  2. Some things are Historic and some things are not
  3. History can be identified as it happens
  4. "Capturing" history - by identifying what is Historic and what is not - is a non-trivial task
  5. the Dispatch performs this non-trivial task every day

Taken together, the premises make an argument for how value inheres in the Dispatch, and why you should buy it. Note that in order to accept the statement, you have to accept the premises, but in order to unpack, you only have to notice something and wonder what makes that thing work. What, one asks, has to be true to make this true, to make this believable? The insight is in the unpacking, the discovery of assumptions.

The insights here have to do with the claims being made about the nature of history: if one wanted, say, to counter-advertise, one could devise an advertising strategy designed around these claims. For instance, one could reverse the time-orientation and claim value in reporting and making sense of the present, and pointing towards the future.

But unpacking doesn't happen in a vacuum: in order to perform this particular instance of unpacking, I had to have cultural knowledge - (that the statement is an advertisement, that advertisements are stories that help sell something, and so on, that advertisements can appear on newspaper dispenser). This is what makes unpacking a particularly ethnographic move: after all, that cultural knowledge had to be acquired somehow. In this instance, advertisements & newspapers are common enough and shared enough that I could do this analysis without having to do research (although you could count my living 5 years in the US as a process of gaining cultural sensitivity; just growing up as a city-bred person counts too, I guess).

In cases where the things being noticed are in less public or familiar contexts - a McDonald's in India, or on the shop floor of an oil rig, or in a hospital - cultural, contextual, and procedural knowledge has to be learned, and often the quality of the learning depends on the amount of exposure one has to that context (what we call 'research'). This is why anthropologists and ethnographers prefer longitudinal participation - the longer one stays in a place, the more one learns, the more things one can unpack and find the premises of, the more powerful the insights. This is also why the ethnographic method has little to do with mere observation alone - you have to do a lot more than that to interpret and find meaning.


The real-time web is producing interesting developments in how information is pushed to a variety of devices and through different channels (twitter, email, RSS, iPhone/Blackberry push notifications). What seems to be missing, however, is an attention to alert redundancy: when the same information is pushed through multiple channels simultaneously, creating more information management interactions than are strictly necessary (e.g. Mint.com alerts pushed to email & iPhone reached my email first, and was seen there before the iPhone notification arrived.

What is needed is a way to mark alerts as 'read' across channels when they are seen on any one channel, and ways to not send notifications after they've been read on any one channel. This, of course, will need the development of a unified notification infrastructure, ecosystem-aware APIs that allow a person's suite of channels and applications to act in concert, models of network performance, and techniques for context-sensitivity that can make judgements about the appropriateness of generating alerts.

Of course, this is a problem that no one organization can solve, but unless it is, we stand in danger of a worse case of information overload than CrackBerry's ever represented.