From the examples in the previous post in the Search series, we have seen that searching is a more constructive activity than punching in search terms and consuming the result. One aspect of this is the ability of the user to quickly scan information looking for characteristics which they can employ to assess relevance. These characteristics are collectively are known as “information scent”.
The definition “those characteristics of information which the user can employ to determine relevance” might sound a bit too vague to be useful, but in reality it invites us to think broadly about what those characteristics might be, and how as designers we might apply them to the information or interfaces we are involved with. It also invites us to consider how the user interacts with these characteristics.
Here’s how information scent might apply to a single search result. When I search for my recent blog posts using the query “information artichoke search”, the big search engines will include a result like this
Upcoming series on search | The Information Artichoke
Sep 27, 2013 – I’ve had an interesting few months advancing my understanding of search from combined information architecture and user experience …
Deconstructing this, I make the following observations:
- The title “Upcoming series on search” – not bad scent, but I’m looking for the series itself, not an announcement. A user however might follow the link to see the pages had links to the series.
- This thought reminds me that I need a table of contents page for the Search series. My other series on Modern Information Architecture has a table of contents and I wonder how it shows up. Not well – its table of contents has the title “Table of Contents”. Not quite what I intended, perhaps “Modern IA Course Table of Contents” would look better here. This is a good example of content that works well in a specific context (i.e. a richly linked blog) not working so well when some context is removed (i.e. in a set of search results)
- My blog name “The Information Artichoke” appears in the title, although I didn’t ask the search provider to do this. Depending on the reader, this might be considered noise or the name of a trusted authority and hence positive scent
- I suspect the search provider knows that this is a blog post
- The date appears in this search result, but not in all
- The description “I’ve had an interesting few months … ” is nicely conversational within a blog, but not high scent here; a short description “We will be exploring such topics as information foraging, information scent, and sensemaking” would be better for a quick assessment.
So do these observations help anyone? Personally, I suspect they could be applied to corporate settings where I am trying to improve enterprise search through the application of metadata and smarter formatting of search results. I decide to give it a try.
I start by abstracting what I have found and make notes to myself, along these lines
- A blog is an example of a type of content. Others might be News, or Reference Material. Coming up with content types to the finest level of granularity is hard work, but some high level typology might be helpful to categorize or refine search results. We might find that not all content needs to get tagged with a content type.
- The name of my blog, The Information Artichoke, might be considered a source or a publisher or origin. Other content types might be tagged as coming from a particular department, or person, or from outside the organization altogther
- The date might be especially valuable for blog postings, news articles, job postings
- The meaning of the date value could be spelled out. For a blog, it could be the date posted; for job postings it could be the date that the job was posted or the date the posting expires; I have an opinion, certainly, but user research will resolve this
- Is there any significance to the date value? We often see a “New” indicator on fresh content, and could imagine other treatments such as red flags on action items
- The description should be constructed to have high information scent. Scraping the first elements of the content does not give good results. For example, the search result description for some of my blog posts include the names of the blog navigation elements; spreadsheets are even worse, showing the contents of a number of cells 3.14, 1.41, 999.
We can desk check our ideas by constructing some sample search results
BLOG POST Upcoming series on search | The Information Artichoke
Posted Sep 27th, 2013
We will be exploring such topics as information foraging, information scent, and sensemaking.
JOB POSTING Senior Information Architect | HR
Closes Nov 5th, 2013 **NEW**
Required to formulate content strategy for sales and marketing
Employee Benefits Overview | HR
Last Updated August 27, 2013
Information on health, insurance, personal development allowance, and …
To proceed, we can stress test the concept on paper, explore business rules for any metadata needed, and talk to our UX colleagues to do usability testing on some of these ideas. They can also suggest some visual treatments and iconography.
Information scent does not apply just to search results, but can apply to documents, pages, links and navigation, among others.
When it comes to links, the simplest implementation is a text link. If the name does not provide high enough scent, then we can add a short description and/or a picture to increase the scent. Another approach is to give the user a chance to lean in for a closer sniff, by providing alt text, or revealing more information when the cursor hovers over the link.
Here’s an example involving documents. I enjoy reading lecture notes in physics and have noticed distinctive aspects of scent that affect my assessment and seem reliable. If the document is all text or all mathematics, it has a low information scent for me. If the math is too hard for me, I rule out the document immediately. Diagrams provide a positive scent. Interestingly, I can make this assessment almost as soon as the document loads in the browser.
I have also noticed characteristics of lecture notes that reflect a distinct bias on my part. If the format is PowerPoint slides, I notice myself downgrading the information scent, but am prepared to persist if the authority is high. If the format is scanned handwritten lecture notes, I rule the document out without further investigation.
This last case is interesting. Most literature on information scent seems to focus on the user selecting stuff to pursue, but this example suggests that information scent is also useful for letting the user select stuff to rule out. Likewise, in our previous examples, somebody not interested in blogs could easily bypass them in a search results page, if they were easily identifiable. This is quite consistent with the biological origins of the concept, which is how animals use scent to identify good foraging opportunities. It also challenges the notion that we want to make content sticky. Maybe traversable might be a better notion.
To summarize, information scent consists of characteristics of information that we can use to make decisions about the value of that information, without having to read the information itself.
I’m not sure about the boundaries of the concept. People are influenced by branding, layout, tone, and a host of other things that we can manipulate. Some people are influenced by US vs. British spellings. Some people discount content with incorrect semi-colon usage. Does this make all of these characteristics part of information scent or something else?
Looking back to the biological origins, we know that bees are attracted to flowers by their shape and colour as well as scent. So maybe we could factor out visual elements. You probably didn’t know that bees can also detect the shape of cells comprising a leaf, and hence whether the leaf will afford their feet a good grip. http://www.botanic.cam.ac.uk/Botanic/Page.aspx?p=27&ix=2847&pi..
Personally, I’m not going to attempt to define the boundaries of information scent. I will just adopt it as a useful concept to have in my toolkit, and apply where it might add value.