There’s a relatively simple formula which describes “weak” or “narrow” artificial intelligence: AI = ML+TD+HITL. To be more specific, this is the definition of supervised machine learning, which is the most common method to produce artificial intelligence. The acronyms in the formula stand for:
- AI = artificial intelligence: in contrast to natural intelligence, is the ability of computer systems to perform tasks or actions that would normally require a human
- ML = machine learning: the ability of computer systems to use algorithms and statistical models to perform tasks without explicit instruction, through patterns and inferences
- TD = training data: the data used to train a machine learning algorithm to perform a task in supervised machine learning
- HITL = the involvement of a human in training a machine learning algorithm
Strong AI – as defined by the Turing test – is when a human has a conversation with a machine and cannot tell it was not a human, based on the way it responds to questions. The optimists believe that strong AI is 10-15 years away whilst the realists/pessimists say not before the end of this century.
How can “weak” AI be applied in research and insights
Over 90% of all human knowledge accumulated since the beginning of time, is unstructured data i.e. text, images, audio and video. The other 10% are numbers in tables which is what quantitative market researchers usually use. Qualitative researchers are the ones using unstructured data, but the volume is limited to a few pages or a few video clips that a person can read/watch in a couple of days.
AI enables us to discover and understand information hidden in large volumes of unstructured data. Machine learning can be used to process text or images in seconds, and annotate them with positive, negative, or neutral sentiment – think sentiment analysis, topics of conversation semantic analysis, and even emotions such as love, sadness, anger and trust. When the unstructured data to be analysed is in text format, the discipline falls under computer science and is called Natural Language Processing (NLP) or Text Analytics.
Semi-supervised, unsupervised- and deep-learning are other forms of machine learning, used to a smaller extent in a market research context, even though deep learning implementation is picking up speed, especially for image analytics.
Use cases for unstructured data analytics
There are a multitude of users, data sources and use cases of data within an organisation. Let’s look at relevant data sources first.
Social media data is a widely used source for businesses in many sectors. Brands from entertainment, fashion, transport, and retail use social media data to engage with their customers regularly.
The discipline that studies social media data is called social intelligence. This discipline revolves around understanding the deeper meaning of what consumers choose to post online, and linking it to a business question. I should clarify that by ‘social media’ I don’t strictly mean social networking platforms, but rather any public source of online data, including review sites, blogs and forums.
Other market research related data sources include answers to open ended questions, transcripts of in-depth interviews and focus group discussions (FGDs), call centre conversations and organic consumer conversations on private online communities.
Beyond market research
Data driven organisations can use text and image analytics to understand unstructured data for the benefit of multiple departments, not just market research for insights. Other use cases include managing brand and corporate reputation (PR), responding to questions and complaints (customer service), discovering and leveraging influencers (marketing), discovering leads who express purchase intent (sales), competitive intelligence, and many more.
Questions that social intelligence can answer
If we agree that social intelligence is currently a widely used application of AI in research and insights then it makes sense to review possible questions that can be answered using it. Some of the questions social intelligence can best answer are:
- Assessing the success of social media advertising in terms of measuring its engagement, reach and positive tone
- Measuring brand performance on social media vs. a competitive set in terms of sentiment and klout (number and quality of followers)
- Better understanding the language used to discuss a brand to inform decisions regarding tone of voice
- Understanding and addressing customer satisfaction problems
Social Intelligence and traditional market research methods
If you are amenable to the statement “social intelligence may replace or supplement some traditional market research methods used to solicit consumer opinions” then here is a list of topics this statement could apply to:
- Advertising campaign tracking
- Brand equity tracking
- New product development research
- Customer satisfaction and experience measurement
- Qualitative landscape framework
Of course, whether social intelligence can just enhance the above, or replace them altogether, depends on the country, language and product category. If you have not embraced the use of AI yet, to tap into the wealth of unstructured data available to us everywhere, then at least keep an open mind and keep asking questions that will help you make an informed decision when the right time comes.
