Why Wolfram Alpha Won’t Work for Historians

In our most recent episode of Digital Campus one of the news items I had a particularly caustic view of was the new search engine Wolfram Alpha. My broader pronouncement in the podcast that WA will “sink like a stone” is predicated on the incredibly clunkiness of the interface and the fact that when the engine doesn’t understand your question or simply has no data to work with, it offers no help…just the statement “Wolfram Alpha isn’t sure what to do with your input.” This alone will send users running back to Google or Yahoo.

But I am sympathetic to the attempt to bring more computational strategies to bear on the search and retrieval of information online. As databases of historical information get larger and larger we are going to need tools like WA (I can’t keep writing Wolfram Alpha) to help us crunch through those databases. So, for instance, I recently wrote something for our 1989 website on the economic causes of the collapse of the communist regimes in Eastern Europe and needed some good old fashioned economic data to work with.

Here’s one reason WA won’t work for historians: Taking the data presented in the source I just linked to, one finds that a 1992 edition of World Bank Facts gives Hungary’s GDP per capita in 1989 as $2,580 (USD). To see what WA comes up with for the same question, I used the query “Hungary gdp per capita in 1989“. When you try this search, the result is a problem. WA offers a different result ($3,097) and a nice graph of Hungary’s GDP per capita between 1970 and the present.

How then can a historian (or student) reconcile the difference between the World Bank’s number and WA’s number? The obvious solution, and the one we teach all of our students, is to check WA’s sources. Here’s what I found–a list of around 30 sources (you’ll have to go to the site and click on Source Information to see them all) with the following disclaimer: “This list is intended as a guide to further sources. The inclusion of an item in this list does not necessarily mean that its content was necessarily used for any Wolfram Alpha result.”

Would you accept a paper from a student with no footnotes but a disclaimer like this one at the top of the bibliography page? No, I didn’t think so. Unfortunately, if my students actually knew that WA existed and I asked them to tell me Hungary’s GDP per capita in 1989, I’m willing to bet the answer I’d get is $3,097 not $2,580. And don’t ven think of asking WA for the GDP per capita or East Germany in 1989. Apparently East Germany never existed and even worse, the GDP per capita result for Germany offers no reference/mention of the fact that East and West Germany merged after 1989.

So, I now have to add WA to my list of websites and web tools to teach my students about in the “these resources have serious problems for historians” category.

4 thoughts on “Why Wolfram Alpha Won’t Work for Historians

  1. Sterling Fluharty

    In its current form, I think you are right. Wolfram has big plans for making this search engine more open, enlisting volunteer data curators, connecting an API, collaborating with the semantic web, etc. This may ameliorate some of the problems you and other historians anticipate. But WA will never work for humanists, in my estimation, until Wolfram abandons or modifies his theory of computational irreducibility in NKS. As I see it, his theoretical bias has already caused him to take a backwards approach to NLP (matching facts to queries, rather than the other way around) and to adopt a vastly simplistic system of ontologies (assuming relationships between objects that are measured in comparable scientific units) in WA.

  2. Pingback: weblog.histnet.ch » Blog Archive » Wolfram Alpha und Geschichte?

  3. Pingback: Suchmaschine | Verdaten. Klassifizieren. Archivieren.

Comments are closed.