3 Reasons Not to Rely Solely on Software for Measurement

June 9th, 2014
by

3 Reasons Not to Rely Solely on Software for Measurement Ellis Friedman BurrellesLuce Fresh Ideas Public Relations PR Media Monitoring Press Clipping News ClippingThe Secret Service is in the market for a social media monitoring and analytics software that can, among other functions, detect sarcasm in social media posts.

Yeah. Because computers are renowned for their wit.

As The Washington Post reported, the Secret Service wants to automate their social media monitoring process and quantify their social media outreach and lots of other things public relations pros and their organizations want.

“More specifically, the orders ask for a long list of specific tools, including the ability to identify social media influencers, analyze data streams in real time, access old Twitter data and use heat maps.”

Is a computer program really the best way to detect sarcasm? Machines don’t have a sense of humor and are notoriously poor at correctly identifying sarcasm and irony; when given a pool of sarcastic tweets, computers identified sarcasm successfully only 65 percent of the time. And that’s when all the tweets were already sarcastic. So why search for a computer program to do a mediocre job when you could hire human analysts to do the job with more accuracy?

Here’s what we know at BurrellesLuce: Software makes an excellent first tier for sorting through data, but there are things computers just can’t do as well as humans. Here are just three reasons that computers are not the pinnacle in measurement:

  • Qualitative analysis improves predictive models: A human will produce more targeted results when going through a random sample of captured data. Those results will improve your predictive models for recurring situations.When you have a set of captured data, a human going through a random selection will produce more targeted results that can be used toward a predictive model for recurring situations.
  • Accuracy through qualitative measurements: When assigning tone, a majority of software programs default to labeling something “neutral” when tone is not clear. If increasing positive coverage is your goal and you know you’re using a program that defaults to neutral, you’re already starting at a deficit and looking at potentially skewed results.
  • Cost vs. Price: If you’re relying solely on a computer sort tone, there will be false positives and false negatives that a human must sort through. The resulting cost in personnel is – more times than not – much higher than outsourcing the original work to a company like BurrellesLuce, which has a dedicated workforce which specializes specifically in gathering and analyzing that data.

So maybe it’s more a question of why are we willing to cede tasks at which computers do not excel to computers? Maybe this is just the next step toward cybernetic revolt. Or maybe it’s time for us to integrate technology with the advanced machines that reside in our heads. CC John Connor.

Leave a Reply