Memex: Perspectives From The User

In an earlier post, I wrote about the opportunity to create a user-centered personal data platform called memex. Memex is both a personal data store and a data platform that enable memex agents, which are personal data-driven applications ranging from personal informatics to behavioral targeting (aka interest-based advertising). This time I will discuss requirements for memex from the user’s perspective.

The user’s memex provides a holistic view of her personal data and thus, as Adina Levin put it, requires a breath-taking amount of trust. So what exactly does the user get in return for this trust? I believe memex agents have the potential to provide the user with dramatically improved computing experiences. The following scenario gives a glimpse into the realms of possibilities.

Uma, the heroine of this story, is a frequent business traveler. Uma just finished meeting with her client and returned to her hotel room in Seattle. She picks up her phone and says “dinner and some music”. The phone “replies” with the following:

A Yelp-ranked list of Thai restaurants and jazz pubs within the radius of 5 miles of Uma’s hotel, some of which have reviews by Uma’s Twitter and Facebook friends.
A special alert informing that Pat Metheny is performing at 7PM tonight at Key Arena center, with ability to buy tickets directly.
An option to tweet-update/invite Uma’s Seattle-area friends to join her for dinner/concert.

With three taps, Uma decides to buy a Metheny ticket, reserve a table at the King and I Thai restaurant around the corner of her hotel, and send the tweet. She then heads to the hotel sauna and massage room for a quick relaxation.

This is the kind of conversational concierge assistant that blends highly personalized search (for restaurant & entertainment), serendipitous discovery (the Metheny concert), novel user interface (speech and gesture), all enhanced with local, social, and contextual information. The crucial piece of technology is Uma’s memex that enables the concierge agent to comprehensively mine her memory extension (dining/music preferences, social connections, location, etc.) for her and on her behalf. Data mining meets personal data mash-up!

Uma cannot remember the last time she googled for dinner and music.

This is not science fiction – I freely admit my lack of imagination and creativity required for sci-fi writing. I simply look at this as something that is tantalizingly close to realization in this 21st century: a digital concierge assistant. How many times have you found yourself on the road and in a hotel room paging through some guidebooks only to give up, order room service, and be content with some junk cable TV?

It’s fun to let our imagination loose in similar sci-fi, but let’s not forget Uma’s more mundane expectations from her memex: safety/security, complete control, device-transparent access, holistic data, and minimal supervision.

Safety and security. This is critical for gaining trust and adoption, and requires state-of-the-art solutions to protect Uma’s data from dangers such as unauthorized access and damage/loss due to cloud failure. We must avoid incidents such as Gdocs security gaffe, AOL’s search data scandal, Magnolia catastrophe, and Sidekick’s almost-catastrophe.
Complete control. Uma is the supreme owner of her memex. She should have email-like control over her memex data such as sharing, deletion, encryption, etc. Some of these operations may have dependencies on individual service providers and thus not always 100% functional (e.g. tweets cannot be deleted), but the platform should aim to provide as much control as possible. Control plays a major role in gaining Uma’s trust and allay her privacy concerns.
Device-transparent experience. This is about providing Uma seamless access to her memex and agents from her devices: her laptop, phone, gaming console, and her browser from an airport’s public computer. Synchronization technologies such as Live Mesh can be used to to build roam-able memex. In addition, Uma’s memex agents are encouraged to be built using an architecture that is mesh-aware.
Holistic data. Amazon.com purchases and wish list, Delicious bookmarks, Facebook contacts and status updates, WordPress blog posts and comments, Twitter tweets, Google searches, etc. Everything should be available at Uma’s fingertip, ideally even when those aforementioned cloud services experience outage. This is the kind of cloud-proof reliability, automated backup and archival that could present a significant freemium business opportunity (e.g. Uma pays for extra cloud storage capacity and automated backup and archival services).
Identities and personae. Just as Uma may have multiple email accounts for different purposes (e.g. work versus personal), she may want to have more than one memex identities as well. Identity technologies such as OpenID and Infocard will play an important role here.
Minimal supervision. The memex and the agents should smoothly run things for Uma with minimal oversight from her. These include managing memex data, negotiating with candidate agents, keep the memex data and agents up to date, etc.

Readers who are familiar with project MyLifeBits at Microsoft Research may have noticed the similarity between it and the current proposal. Think about my memex proposal as MyLifeBits that is optimized for mining and mash-up to improve user productivity and marketing efficiency, and perhaps less emphasis on issues such as long-term archival and re-encountering.

Marketing efficiency? Yes, you read that right. I believe it’s time to let loose the genie that is online advertising from the bottle of covert operation, privacy struggles, and data silos. Open and transparent memexial targeting done right will benefit everyone: consumers, advertisers, publishers, and John Wanamaker.

So which companies are vying to build memex for Uma? That’s the topic of my next post.