IBM Overthinking It with A “Cognitive Dress”

Earlier this year IBM collaborated with New York fashion designer Marchesa to produce a tech-infused dress for the Met Gala. The collaboration involved using the IBM Watson supercomputer to assist in the design of what they dubbed a “cognitive dress”.

Technologically, the dress itself used materials and LEDs that change colors under digital control. The “cognitive” aspect apparently is that they use IBM’s data analytics to derive “sentiment” about the dress—from tweets about the event. IBM also has an AI that supposed understands color, which is used to correlate the tweets to the color of the dress. In other words, this is a sort of data visualization system, using the dress as the display.

The Watson AI also contributed design concepts, at least partly based on automated learning of the Marchesa “style”, which created suggestions to the designers. I’m imagining that Watson was creating a Marchesa design that is “more Marchesa than any other Marchesa ever was”, or something like that.


My first reaction to this project is, “what in the world would IBM Watson have to do with this?” As I read the description, I wondered, “why?” and “who cares?”

First of all, stripped of the hype, the actual computational features are rather pedestrian. I also had a bit of trouble figuring out what is “cognitive” about this dress. The technology is all about emotions, not cognition of any kind. Eventually I figured out that IBM uses “cognitive” as a tradename for anything that the Watson computer does.   Sigh.

Anyway, I’m sure Watson does a good job at automated learning, but automated learning is scarcely a new or untried technology. Sentiment analysis, color selection, and even style advice have certainly been done before. Tying the output to wearables is definitely not new.

Low marks for technical novelty and difficulty.

Second, the specific analytics are unimpressive. Does Marchesa need a hundred million dollar computer to advise them on style and color? Who cares what the general “sentiment” about the Met Gala is on twitter? What does that even mean, and why would I want to know? Mapping this alleged “sentiment” to allegedly representative colors is, well, silly.

However spiffy these technologies, this particular project is a pointless use of them.


From the point of view of the dress itself, I have to say that the technology is misguided. Deeply misguided.

First, a fashionable frock should highlight and enhance the wearer and her companions. “Look at me”, the dress should say, “notice me, notice my dress”.  “Admire me, desire me”.  Etc.  The message should be about the woman.

But this dress is driven by data that is semantically and logically unrelated to the dress or its wearer. This is basically a walking signboard, sending messages about the crowd and, I’m afraid, IBM.

Not only is the message from the dress irrelevant, it actually distracts from the “look at me” message. To the degree that you notice the dress and its tricks, you are drawn to think about flashy lights, twitter, IBM, and the Internet, obscuring the beautiful and charming woman. This is exactly what you don’t want fashion to do.

I also find the entire notion of linking a wearable to the Internet to be misguided. If I’m going to the Met Gala (which isn’t terribly likely), I want to be there, paying attention to the event and the people there. Why would I want to be paying attention to the chatter on twitter or to “to join the cognitive conversation on the red carpet”? This distracts from interacting with the people in the room, distorts and misrepresents the experience.


Finally, the “design assistant” aspect is less absurd, though I really wonder whether IBM Watson had much to teach Marchesa about designing dresses. I would say that the idea of advising an artist on new work based on their previous work is a slippery slope. How will the designer innovate, if the advisor is helping her to “copy herself”? I think designers should be very skeptical of this concept.


Overall, this project seems like a misguided use of Watson’s capabilities, and certainly not a great idea for what “smart fashion” might be.

(Hint: I want a “smart” dress to do clever things to promote the messages of the woman wearing it. Things that say “look at me, desire me, envy me” in subtle and powerful ways, ways that might be tailored for each viewer. And things to say “come closer” to the right people, and “keep away” to the others. )

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s