In this session, Ray sets the scene of where AI and Automation are currently being used in and around market research, the key terms and concepts, the main successes and areas of conflict.
3. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
Automa+on
• Glass blowing automated New York
1905 for boGle produc+on.
• BoGles per person up 600%
• Price per boGle fell to 8% of previous
price
• Many skilled glass blowers lost their jobs
• Reduced costs and increased consistency led to new uses for glass boGles,
including foods, drinks and medicines
• Massive increase in the use of glass, lots of new jobs, economic growth, health
improvements (because food was safer)
8. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
Automa+on & Market Research
• Herman Hollerith, US Census 1890
• 1970s and 80s
– Op+cal scanning of surveys, CATI, CAPI, SPSS, spreadsheets,
desktop publishing, presenta+on & char+ng so]ware
• 1990s & 2000s
– Online surveys, CAQDAS, DIY surveys, Online access panels,
dashboards and reportals
• 2010s
– ‘Shrink-wrapped’ research, chatbots, text & sen+ment analysis,
image and video analysis, report automa+on, advanced analysis
e.g. semio+cs and conjoint analysis
10. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
Automa+on and Jobs
Automa+on has already cost hundreds of thousands of jobs in MR
– Interviewers, data punchers, call centre staff etc
Automa+on will change exis+ng projects and jobs
– Faster, cheaper, more standardised, and needing fewer people hours
– Many jobs will go, and some new jobs will be created
• E.g. ‘clerical’ jobs will decline, storytelling and client success managers will
grow
If the market grows, there will be a growth in jobs
– Even in clerical jobs, but the +me spent on each project will be much
shorter than before, so more projects will be handled and the role will
be more Q&A, interpreta+on, sales etc.
12. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
AI
Ar+ficial Intelligence Machine intelligence
Machine learning
Expert systems
Bots
Intelligent agents
Neural networks
Inference engines
Natural language processing
Deep learning
Evolu+onary algorithms
Bayesian networks
Anything ‘clever’ J
Alexa, Siri, Google, Home
Autonomous cars
Game playing – chess, Go etc
Chatbots
Programma+c adver+sing & marke+ng
Predic+ve analy+cs
Voice recogni+on
Text analy+cs
Image analy+cs
Transla+on so]ware
14. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
Algorithms, Rules, Expert Systems
Strategy for playing Noughts and Crosses
1. If opponent has two in a row, take the remaining square. Else
2. If a move "forks" to create two sets of two in a row, play that move. Else
3. Take the centre square if it is free. Else
4. If opponent has played in a corner, take the opposite corner. Else
5. Take an empty corner if one exists. Else
6. Take any empty square.
This used to be the main way to do AI, but now ‘machine learning’ is
more prevalent.
Many MR applica+ons use/will use expert systems, e.g. how to design a
project and reports from data
15. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
Learning Systems
Supervised Learning
– Given a set of inputs and outputs, the machine learns
to to generate the outputs from the inputs
• For example, show it 1000 coded open-ended comments
and ask it to do the same to 1 million
Unsupervised Learning
– The machine looks for paGerns and structures in the
data
• Cluster analysis is a simple example of this
16. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
Deep Learning
• Also called hierarchical learning
• Uses mul+ple levels, where the outputs of one
level feed into another.
• For example,
– Teach the machine to play chess, give it some ini+al
strategies (e.g. based on real games) and the ability to
be random
– Let it play itself over and over, learning the best
strategies
17. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
AI and Market Research
• Chatbots and adap+ve surveys
– Models built for non-MR use, tweaked for MR
• Image, text, and video analy+cs
– Including coding and biometrics (e.g. facial coding)
– Supervised learning, either from MR or elsewhere
• Predic+ve analy+cs, aGribu+on analy+cs, mining meaning
– Supervised learning and unsupervised learning
• Report wri+ng / story finding
– Expert systems and unsupervised learning
• Project design & sample management
– Expert systems ini+ally, later supervised analy+cs
18. An Introduc+on to AI and Automa+on
Ray Poynter, NewMR
The Future of AI
& Automation
AI and MR in Summary
1. Most things that are ‘clever’ will be called AI
2. Most developments will happen outside MR and then be
u+lised – e.g. AWS
3. Machine learning will focus on situa+ons where learning
sets exist – e.g. coding
4. AI will allow materials that have tradi+onally been
analysed with qual to be quan+ta+vely analysed
5. Expert systems will be core to anything that does not
produce large training sets – e.g. project design