features archive | main home

Social Software ideas

December 30, 2003

While social software may be the internet revolution du jour among venture capitalists, as a user I'm still waiting for the killer social software app that lives up to all the market hype. Recently I've been thinking about how the current crop of options could be improved upon, or at the very least, how they could be leveraged to be something useful for users. I've come up with a few ideas, some half-baked, others fully baked. I offer them here in the hopes that someone, somewhere already built it or would like to build it.


Colloquial mapping

The gist
Yahoo Maps + Slashdot

Have you ever been literally steered wrong by automated mapping systems like Yahoo Maps or Mapquest? Either the maps are out of date or the algorithm that determines the shortest distance between two points doesn't account for local traffic patterns. I find the difference between how Yahoo Maps says I should get from a point A to a point B and how a friend or family member would tell me how to get there differs about half the time. I find that the more rural or off-the-beaten-path a destination is, the bigger the difference.

I'm finding that in Oregon, the speed difference between two lane roads with stop signs and four lane freeways is substantial. Yahoo Maps suggests the shortest distance and it almost always includes backroads that are riddled with delays. Where these systems fall short is that they are not aware that a quick 5 minute jog over to a major freeway can save 20 minutes or more on an hour long drive.

The idea
Create a mapping service that allows the community to suggest alternate paths along with reasons for it. Suggest the routes to be taken as dictated by the software's GIS information, but also list user suggestions, and to ensure quality, also add ratings of user suggestions by others.

If I wanted to go from San Diego to Ventura, CA, there are several paths I could take and it would depend on time of day, day of week, time of year, and/or the weather. Software would dictate that I go directly up the 5 and 405 freeway to the 101, but that'd be murder during rush hour on a weekday and might take 4-6 hours to cut through three major metropolitan areas. During a holiday period, you'd probably save lots of time taking an eastern route around most of Southern California even though it would be a longer distance.

The rub
The biggest obstacle I can see is normalizing all the data. By its very nature, getting directions from people would be "dirty" and require some significant logic to normalize and get it into a database format that could be queried. You could store the paths based on the geographic start and end points, and perhaps take long/lat points along the way, then you could show users paths suggested by users that had start and end points that were shorter or longer than their desired path. The trickiest thing would be providing user suggested paths for a trip between say Los Angeles and Ventura, CA that could correctly draw upon some of the suggested paths from the San Diego to Ventura submissions.

Allowing other users to rate the quality of suggested directions would hopefully keep bad directions to a minimum and at the same time float the best alternate paths to the top. With proper reputation management in place, the service could keep track of a user's overall quality of suggested directions and highlight those ingenious travelers that always seem to know the quickest way to get somewhere.

Along with the proven utility of internet mapping, adding a suggested route system could fill in the last remaining gaps and produce a hybrid automated and human created system that any amount of AI programming couldn't match. Mapping software relies on simple mathematics and a conceptual map of the earth's surface that imaginary vectors can be plotted along. Humans that are veteran drivers in a particular region have extensive knowledge based on years of experience that simple mathematical models can't replicate. This service would attempt to bring those two worlds together to provide the best possible experience for users trying to get from point A to point B.


2. Geographical opinion systems

The gist
Epinions + Friendster

Last summer I moved to a town in a place far away from where I've spent the past few years, and one of the first problems I had to solve was finding the perfect everything. I quickly amassed a bunch of questions that took months of trial and error to answer through a network of new friends and neighbors. Where could I get a good haircut? Which one of the local dentists would be most understanding of my dental anxiety? Which store should I shop for food at if I want a lot of organic, natural, and meatless food? Are there any trustworthy mechanics in this town? Which one of the two Thai places is "the good one?" Where should I go for a nice night out here? Which theater plays the art house movies? Which one of the furniture stores should I trust with my money?

The idea
Even with a small network of friends, it's tough to find answers for all the questions you might have when you move to a new city. Worse yet is traveling to a new place and having to send off emails to friends that live there, asking them for all their favorite places to eat and have fun. The crux of this idea is to build something that combines a service like Friendster with a review site like Epinions. Basically the site would serve as a digital representation of the connections and knowledge one builds over time when living in the same place.

I'd be more than happy to write detailed reviews of all my favorite places in San Francisco for friends. Currently I do it over email when asked by friends and on localized or private email lists when the questions come up. It took me years to find the one honest mechanic, the nicest dentist on earth, and the best sushi in San Francisco. I share knowledge freely with my friends and they do the same right back at me, but it's tough to keep track of this stuff. This site/service would serve a place to share this information among a trusted network of friends or strangers given appropriate levels of privacy control and reputation management.

The rub
The hardest part of course is weeding out the tainted information. If limited to your friends network, this wouldn't be too hard to manage, as you probably do it currently. I might not treat my friend Jonah's opinion of hairdressers highly because I know he is married to one and talks up his wife's business too much. The bigger rub is when you expose the reviews to strangers, because it quickly becomes a minefield where trust management is paramount to keeping the service useful. Given any system that tracks quality or quantity of any property, participants will game the system to rise in rankings. People signing up fake accounts to rate their own services highly would be the death keel for this service among the many possibilities for tainting data.

Another obvious problem is why anyone would enter all this information somewhere. Sure, epinons could someday do this and it was fun at first, but writing reviews takes work, and I believe epinions relied on encyclopedic reviews and should have instead allowed shorter 1 paragraph reviews that might only be meant for friends. Reducing some of the formality and perceived amount of work would reduce the friction of getting information. Using perhaps an existing network like Friendster, keeping reviews private to only those close to you would also help. If you wanted to share a review of your favorite bar, it wouldn't mean spending fifteen minutes writing to help out some nameless corporate website network, if it was for your friends only, then you'd have the more immediate feeling of helping people you care about out. This would also give an incentive to expand your friend networks, in order to access to this privelaged, private information.

Tribe.net sort of does this by putting some emphasis on geographical location, though they seem to be more of a Craigslist style classified ad network. When I log in, I usually see a list of strangers trying to sell a guitar amp or get a knitting group together, which is fairly different than what I envision here.


3. Collaborative consumed media

The gist
Friendster + ??? (some sort of media management service)

Friendster currently lets you list interests in a free-form way and those become links to others interested in the same subject matter. What I want here is something a bit more formal than "rock music" or "AC/DC". I'd love to know the last five books my friends purchased and the last five CDs they liked. I'd love to know what's spinning in their MP3 player currently and what DVDs they enjoyed watching recently. While this may be a potential privacy problem, as long as my friends are the only ones that can see this info, I wouldn't mind sharing it, and I'd love to check up on what media they've been loving recently.

The idea
A central service perhaps built upon existing systems at Amazon or Friendster that allows you to share as easily as possible all the media you are consuming. Currently my friends do this using a mish-mash of web services, spaghetti code scripts, and their weblogs. I might think about getting the New Pornographers album after I hear my friend Andre rave about it on his site, or if I see Jason's "now playing list" that features the tracks.

The rub
As with the last idea, reducing the friction as much as possible is key to allowing people to enter data into the system. At amazon, simply let people share their recent purchases with friends and perhaps let them write micro-reviews that don't show up as formal amazon reviews. Provide javascript bookmarklets to let users quickly add movie listings, book ordering pages, and band homepages into their media library to share with friends. This would be trivial to tie into an existing commerce system like amazon, which already has the data and only needs to make it easy as possible for me to say "hey friends, I just finished reading Word Freak and it was a blast from start to finish!"

Privacy concerns would have to be addressed, as this could be a Total (media) Information Awareness listing of everything you've bought, watched, listened to, and read. Users would have to trust the company running the service and trust that their data wouldn't be used against them in any way.

Currently Erik's ingenious service All Consuming does this by scraping weblogs (and letting you dictate which ones are your friends) and creating pages and sending you alerts when friends mention new books, though it doesn't do music and movies, and it sort of does a end-run around the data by grabbing it from your site and normalizing it into an amazon-like framework.

Audioscrobbler does "what's playing" lists pretty well (here's my out-dated profile), perhaps if they could be incorporated into an amazon purchase history, it could be extended to movies and books.

4. Reputation management ideas

The gist
Multi-variate reputation management

Reputation systems have been around for a few years and it's about time to improve upon them. I've noticed that after using eBay for a few months that the simple +/- rating system doesn't always tell you what you want to know about an unknown seller, nor does it equate with trust in some cases. This could work with any reputation system, but eBay comes to mind as the most obvious application.

The idea
This is less of a product or service and more like a bunch of ideas around reputation systems.

Add additional variables for tracking reputation in a community system. For eBay, there are numerous informative data points that could help calculate trustworthiness. I'm going to trust someone that has high ranking for selling ten $1,000+ items more than someone that sold 50 $5 trinkets, and doubly so if I want to buy a $500 item. After having been burned by an overseas transaction gone wrong, I'm wary of buying from sellers that are very far away, unless they have extremely high marks.

Similar to the "grade inflation" problems common in colleges everywhere, everyone at eBay seems to have the same positive rating of "GOOD COMMUNICATION WOULD USE AGAIN A+++++++++++" which doesn't really tell you much and becomes meaningless if every bit of feedback looks like that. If you plumb someone's profile and all their previous transactions, you can eventually figure out if they've moved any big ticket items but it'd be great if that information (which is already in the database) could be surfaced and used to assess an adjusted reputation ("This user has been rated positively on 125 sales of items over $500". Other bits of info could include things like location of both the seller and the buyers (I'll trust a seller in Spain if they have lots of high ranking sales to others in my country).

Besides eBay, a system such as the one in use at Slashdot that keeps tally of a user's karma could be expanded to include additional datapoints, such as # of words per highly rated answer. You might prefer encyclopedic answers, or you might instead like to know the efficient users that pack the highest ratings in the fewest words/post.

Additional data would really help out sites that involve larger sums of money changing hands. The eBay-like freelance job site eLance and the home improvement finder site Service Magic could improve their simple +/- systems with a cost calculation. I wouldn't hire a handyman that typically does $50 doorjamb fixes to hang $10,000 worth of windows and at eLance, if I was looking for a cheap $500 website for my business I'd hire someone that had successfully done those jobs in the past and had a portfolio filled with low-cost jobs.

Update: Jay Allen lists an idea I would also love to see, an address book that could be shared with friends. This past xmas I had to email my new address out to a couple dozen people that asked, and most all of them were already connected to me in friendster.

 

Creative Commons License As with all Features on this site, the image and essay can be freely copied and redistributed less the restrictions outlined in the attribution noncommercial sharealike license.

features archive | main home