Brynn Marie Evans

Social search behavior is now generally agreed to be common enough that companies like Google are buying up companies like Aardvark, and academic researchers are asking good questions about the value of networks for question-answering (Q-A). I have done a bit of research in this area myself, and so I was quite pleased to read about this new study by Microsoft researchers Merrie Morris, Jamie Teevan, and Katrina Panovich.

They were building upon (among other things) my work on asking and answering behavior on social networks (“Do your friends make you smarter“) and wanted to know more details about things like: why do people post questions, answer questions, and what motivates people to go the social search route over traditional search.

I want to talk about their conclusion here first: where they discuss the tradeoffs between using social networks and search engines. This was the most interesting part of the paper and is the part that has the most practical take-aways for the UX community.

Basically, for questions requiring trustful and personalized responses like for recommendations or opinions, social networks were strongly preferred over search engines. People occasionally perceived a delay in response time (but sometimes did not!) — and either way, this delay was not a downer since it saved people the effort of making multiple attempts on a search engine before getting the right answer back (if they ever did). Also, subjective questions got faster responses on social networks than objective questions!

Additionally, they talk about how being on social networks is fun for people. It’s too bad that they under-emphasize this point because it’s really important from a social interaction design perspective! People engage on social networks because it’s fun! There’s an emotional tie to the community that keeps bringing them back. When it comes time to ask a question, people turn to their friends for help because it keeps them connected with others, helps them to share information about themselves, and keeps them engaged in their community. Fun is an important element of design — and until Google, Facebook, Yahoo, Microsoft, or other search providers make the information seeking part fun, they will lose. It’s not enough to simply pipe in real-time Twitter results and expect that to be fun or more personalized!

That (above) is the juiciest part of the paper, but I do summarize the whole paper below for other nuggets of inspiration.

To start, the authors note that Q-A on social networks is unique (like unique from Q&A sites like Yahoo Answers) because:

• questions are posted with true identities (rarely anonymously)
• people’s audience is smaller (being only people’s direct contacts, or possibly friends-of-friends)
• status messages are short: so questions are succinct

Then they collected surveys from 624 Microsoft employees (1/4 were female) about their Q-A behavior on Facebook and Twitter mostly. The survey asked about:

Asking questions:

• whether they ever used their status messages to ask questions
• if yes, to share a recent example of a question they asked
• what kind of responses they received
• how often they logged onto social networking sites

And answering questions:

• whether they responded to someone else’s question (from a status update)
• if yes, to share a recent example of a question they answered

50% of people who replied to the survey did indeed use their status messages to ask questions! Sometimes they were silly questions (“Why are men so stupid?”) and sometimes they were serious (“Point and shoot camera just died — need to replace it asap. What should I buy? Think under $200″).

On asking questions

Questions took many shapes and forms. Some were targeted to “Anyone” in the social network; others were general statements put out to the network to interpret. Most questions were asking about recommendations (29%) or opinions (22%). Some were about facts (17%) and some were rhetorical (14%).

Questions also related to everything from technology (29%) to entertainment (17%), places (8%), current events (5%), and ethics and philosophy (2%). People said they’d generally avoid asking their network about topics like health, pornography, religion, and financial issues because those are too personal.

Another finding was that Twitter users were more likely to ask questions about technology, while Facebook users more likely to ask about home and family issues. This isn’t too surprising, but does show that the makeup of people’s networks affects what kinds of questions people will ask.

One looming question in this research is: why do people choose to ask their friends over searching on Google? In this study, people reported having more trust in their friends’ responses or that they thought traditional search wouldn’t work for them with this particular question. Sometimes it was obvious that their friends would know their history, family situation, and other preferences better than a search engine.

People also generally knew the makeup of their networks (meaning they could get targeted responses) and people wanted to advertise their current interests by way of asking questions of their friends.

What about answer speed or answer quality? There were mixed results, but often questions asked on social networks were not urgent so any delay was considered acceptable. It was also easier than sorting through tons of results on search engines and quite often more fun to engage with friends.

On answering questions

Almost everyone has answered a question on one occasion or another. And of the questions offered up in this survey — nearly every one received an answer! That’s kind of amazing considering Q-A sites in general have a much lower response rate, but it’s also likely due to the intimate make up of people’s social networks. Not to be an under-appreciated finding in this paper, however!

About one-quarter of questions got an answer in 30 minutes or less, and 90% were answered within one day. (I’d LOVE to see how Aardvark’s data compares here. I have several questions out to Aardvark that have never been answered; but of the ones that do get answered, it’s typically faster than 30 minutes.)

Yet people don’t seemed to be bothered by the 1-day response lag. Perhaps this is because replies from trustworthy friends are more valuable than junk search engine or Yahoo Answers responses?

Why did people respond to questions by others? Most often they were just trying to be helpful (37%) or had expertise in the area (32%). Those seem like good reasons to me, and ones that are often overlooked when thinking about how valuable personal social networks can be to people. At other times, people replied because they had a good relationship with the asker, were connected socially, or had some notion of social currency (that answering now may result in a favor later).

When people didn’t respond to questions by others, it was mostly because they didn’t know the answer! That’s pretty neat. But some said they’d prefer a personal request and ignored questions that were asked to the network as a whole. In fact, my research on Q-A in social networks revealed the very same thing — that people gave great responses when asked privately or one-on-one, even when they weren’t very knowledgeable about the topic.

Another interesting, but not very surprising finding, is that people who were more frequent social networks users received quicker responses! Why duh! People who are engaged with their communities have developed rapports and back-and-forths and social capital with folks — what goes around comes around. But it is cool to see this intuition confirmed in real data!

Unfortunately, this paper is weak on design recommendations. My colleagues and I have already suggested the very same ideas (e.g., surface your friends status messages on the results page — what I call “friend-filtered search” — or have a search engine post a question as a status message on your behalf.) Even as I say this, there are serious social interaction design issues with this approach. Expectations about search engine behavior, expectations about social network intrusion, privacy (obviously), interactions with online agents, etc. all come into play. I very much hope that these issues are being considered by the big search engine companies, but somehow I’m afraid they’re not.