How to spot a good A.I. tool
Share on facebook
Share on twitter
Share on linkedin

Advances in A.I. over the last few years have reached exciting new developments when it comes to delving into the human psyche. We at Jade Kite strongly believe that now is the time to integrate machine learning into research methodologies for replicable data about human emotion.


As with all new technologies however, we recognize that it can be hard to know where to start when looking for a new partner. It is vital to be able to discern which tools produce the most accurate data and add value to existing approaches, rather than derailing them.


Having studied A.I. and followed its progression for over 20 years, we understand exactly which questions to ask when considering a new partner and what to look out for. Here are three key areas to consider.


  1. Data Input

A.I. is only has good as the data it has been trained with, and neural networks (that replace “hard coding” in modern AI) need both quantity and quality to produce cohesive and reliable results. Look out for large datasets with millions (or preferably billions) of touchpoints. Be wary about data that comes solely from internet scraping and be sure to check the sources; using only posts on Facebook or Twitter for example will not usually not give the depth of information required to accurately train an A.I. where human emotion is concerned. Instead look for data that uses transcripts and past research as part of the data input. Carefully assess the data annotation and characterization processes to avoid inherent bias.


  1. Machine Learning

Education of these tools should be continuous and ongoing to ensure that A.I. uses relevant data to keep growing in knowledge. This requires human input to check the data and make adjustments. While you want to be avoid teams that constantly correct a large amount of inaccurate results, you also need to accept that a certain amount is absolutely necessary. We engage with tools that use human scrutinization across 1-2% of the data as continuous quality control.



  1. Accuracy

A.I. will always produce outputs, but accuracy levels vary across systems. Any developer will be aware of the percentage level of accuracy. In order to engage with tools in a meaningful way, it is important that this percentage level remain high. Asking this question upfront (and gauging the transparency of the response) will give an immediate indication of how useful a specific tool is at this given moment.


Constantly scanning for the latest A.I. technology and assessing their efficacy and potential for qualitative research can be timely and costly. Jade Kite is constantly scouting and testing the latest tools to pull together a top tier toolkit of the latest technology. Get in touch today to find out more.


Ready to kick-off your project?

If you’ve got a problem that is proving tough to crack with your usual avenues, you’ve come to the right place.


12 Responses

  1. I think tis iss one oof the most significant information foor
    me. Annd i’m glawd readding your article. Butt should remark
    on ffew genesral things, The site style is ideal, the articles
    iss really excellent : D. Good job, cheers

Leave a Reply

Your email address will not be published. Required fields are marked *