Conference Notes, 21st century common sense: Using collective intelligence to tackle complex social challenges, Oct 16 2019
What’s more intelligent, people or machines?
Don’t get me wrong, machine learning and artificial intelligence are fascinating. A combination of statistical, computational and mathematical methods — able to identify patterns, simulate understanding and even transfer learning from one domain to another.
… but so are people! Humans are just really good at having a life, lived experience, domain expertise, being part of the world and knowing about context. I would argue that what makes any system intelligent is often somehow a human element. Engineers that fine tune and maintain a system, a power user and domain experts that feed expertise into the system, or contributors who act as eyes and ears or react to a system.
What if … we wouldn’t have to choose?
“A simple recurrent question: How can we mobilise more intelligence — data, insights, ideas — to solve problems?”- Geoff Mulgan (Chief Executive, NESTA)
Collective Intelligence (CI) describe the idea to combine human and machine intelligence and consider both when designing systems that take into consideration context, prior experience and the ability to abstract, unlearn, relearn and react. By keeping the human-in-the-loop with computational systems, we can use artificial intelligence to facilitate “wisdom of the crowds” and design systems that scale beyond what is possible by either humans or machines only.
- Understand Problems Generate contextualised insights, facts and information on the dynamics of a situation.
- Seek SolutionsFind novel approaches or tested solutions from elsewhere. Or incentivise innovators to create new ways of tackling the problem.
- Decide and ActMake decisions with, or informed by, collaborative input from a wide range of people and/or relevant experts.
- Learn and AdaptMonitor the implementation of initiatives by involving citizens in generating data, and share knowledge to improve the ability of others.
Last week NESTA’s Centre for Collective Intelligence brought together researchers from various disciplines to share ideas and findings in London, UK. Panel discussions and talks shared a number of projects that apply CI to solve system challenges.
Breadline | NGO volunteers use an interactive collaborative map to find left over goods from bakeries to plan an optimal route and collect food for redistribution. The found Daisy Tam is researching food security and the use of such tools to alleviate its challenges.
OSR4Rights | In this research project led by Yvonne McDermott Rees (Swansee University) human rights violations are documented through collective intelligence. Huge volumes of recordings on mobile phones allow to collect evidence, yet identification of violations is challenging.
Read more: https://osr4rights.org
Many more case studies can be found here: https://www.nesta.org.uk/feature/ai-and-collective-intelligence-case-studies/
A panel on designing AI that extends human capabilitiesbrought up some interesting thoughts.
Authorship | Deep learning “makes all past authors part of the collective intelligence [or authorship]” says Marco Marchesi referring to art generate based on historical art data sets.
Triage | Julien Cornebise talked through case studies in health or emergency situations where automated assistants can help triage urgent cases from less urgent cases — but leave the actual assessment and decision making to the human expert
Read more: http://www.cornebise.com/julien/
Boredom and creativity | Karina Vold argued that by striving for ever more efficiency and by removing “boring” and basic repetitive tasks, we potentially also eliminate time to think, reflect and serendipitously discover short cuts.
Read more: https://www.kkvd.comhttps://www.kkvd.com
Bias | A brilliant example going “to far” was presented in the “Is There a perfect mum” advertisement. The deliberate campain showed a portrait of an “ideal” mum based on a biased data set, containing blonde, white female models only.
Alternative Data Sources | In the attempt to tackle emergency situations such as Forest and peatland fires in data poor areas Maesy Angeline (Pulse Lab Jakarta) described the approach to use mobile phone connection data or social media (e.g. instagram) in combination ethnographers on the ground to better understand the impact of the disaster.
- https://youtu.be/pR4QVX_eisI, Haze Gazer, Pulse Lab Jakarta
Creative Play | Artists showcased their studios’ work on engaging with space and artificial objects (Florian Ortkrass, RANDOM INTERNATIONAL), engaging with public space and cities (Usman Haque, https://umbrellium.co.uk) and 3D representations (Kadine James, www.hobs3d.com). A fun case study showed how kids could interact with a Building Management System and learn about the status and sensors of their college building.
- https://youtu.be/Ik2bogIsuUQ, Augmented BMS, Umbrellium
What’s in for us?
Collective Intelligence, Systems Thinking, Service Design, all these movements have one idea in common. That is, to design computational touch points as well as to consider human expertise and define human roles that input in or react as core part of the system.
The tools we provide are never used in isolation. They are always part of an ecosystem, used in combination with various other applications and services and based some output or used in order to input to another system. At the same time, many of the services are provided based on human domain experts input. Being more aware of this emerging ‘Collective Intelligence’ certainly could help us make even better product decisions!
Getting Started …
Whenever thinking about an intervention (PoC, new service, process change etc.) we should ask ourselves the following questions:
- What are all factors that contribute to a system’s intelligence?
- Which (machine) team (algorithm working with data experience) are we putting on which tasks?
- What data set, aka experience, are we hiring?
- What are biases within that experience?
- Which tools and algorithms are we hiring?
- Which (human) team are we putting on which tasks?
- Who contributes content?
- Who contributes evaluation/rating/training?
- Who reacts to other actors output?