What can we Learn from Artificial Intelligence?

This might seem strange, suggesting that a science dedicated to learning from how we humans operate, could actually return the favour by teaching us about ourselves? As strange as this may sound, this is precisely what I am suggesting.

Having spent a good deal of my early career in the “first wave of AI” I had developed a healthy scepticism of many of the capability claims for AI. From the decade or more I spent as an AI researcher and developer I had come to the conclusion that AI worked best when the domains of endeavour were contained within discrete and well bounded ‘solution spaces’. In other words, despite the sophistication of mathematical techniques developed for dealing with uncertainty, AI was simply not that good in the “grey” areas.

AI’s Second Wave


The “second wave of AI” received a big boost when Google company Deep Mind managed to up the ante on IBM’s chess playing Deep Blue  by defeating the world Go champion Lee Sedol. According to Founder and CEO of Deep Mind Demis Hassabisis,  the success of their program AlphaGo could be attributed to the deeper learning capabilities built into the program, as opposed to Deep Blue’s largely brute force searching approach. Hassabisis emphasizes the ‘greyness’ in the game of Go, as compared to Chess. For those familiar with this ancient Chinese game, unlike chess, it has almost a spiritual dimension. I can vividly recall a research colleague of mine, who happened to be a Go master, teaching a novice colleague the game in a lunchtime session, and chastising him for what he called a “disrespectful move”. So AplhaGo’s success is indeed a leap forward for AI in conquering “grey”.

So what is this “deep learning” all about? You can certainly get tied up in a lot of academic rhetoric if you Google this, but for me it’s simply about learning from examples. The two critical requirements are the availability of lots of examples to learn from, and the development of what we call an “evaluation function”, i.e. something that can assess and rate an action we are considering on taking. The ‘secret sauce’ in AlphaGo is definitely the evaluation function. It has to be sophisticated enough be able to look many moves ahead and assess many competitive scenarios before evaluating its own next move. But this evaluation function, which takes the form of a neural network, has the benefit of being trained on thousands of examples drawn from online Go gaming sites, where the final result is known.

Deep Learning in Business


We can see many similarities to this context in business. For example, the law profession is founded on precedents, where there are libraries of cases available, for which the final result is known.  Our business schools regularly educate their students by working through case studies and connecting them to the underlying theories. Business improvement programs are founded on prior experience or business cases from which to learn. AI researchers have taken a lead from this and built machine learning techniques into their algorithms. An early technique that we had some success with is called “Case Based Reasoning”. Using this approach, it wasn’t necessary to articulate all the possible solution paths, which in most business scenarios, is infeasible.  All we needed to have was sufficient prior example cases to search through, to provide the cases that most matched the current context, leaving the human user to fill any gaps.

The Student Becomes the Teacher

Now back to my question; what can AI now teach us about ourselves? Perhaps the most vivid learnings are contained in the reflections of the Go champions that AlphaGo had defeated. The common theme was that AlphaGo was making many unconventional moves, that only appeared sensible in hindsight. Lee Sedol has stated his personal learning from his 4-1 defeat by AlphaGo in these comments: “My thoughts have become more flexible after the game with AlphaGo, I have a lot of ideas, so I expect good results” and “I decided to more accurately predict the next move instead of depending on my intuition”. So the teacher has now become the student!

It is common for us as human beings to become subjects of unconscious bias. We see what is being promoted as a “best practice”, perhaps reinforced by a selected few of our own personal experiences, and are then willing to swear by it as the “right” thing to do. We forget that there may be hundreds or even thousands of contrary cases that could prove us wrong, but we stubbornly stick to our original theses. Computers don’t suffer from these very human traits. What’s more they have the patience to trawl through thousands of cases to fine tune their learnings. So in summary, what can we learn from AI?

  • Remember that a handful of cases is not a justification for developing hard and fast rules;
  • Before you discount a ‘left field’ suggestion, try to understand the experience base that it is coming from. Do they have experiences and insights that are beyond those of your own close network?
  • Don’t be afraid to “push the envelope” on your own decision making, but be sure to treat each result, good or bad, as contributing to your own growing expertise; and
  • Push yourself to work in increasingly greyer areas. Despite the success of AlphaGo, it is still a game, with artificial rules and boundaries. Humans are still better at doing the grey stuff!





SWOOP Video Blog 2 – Yammer Groups

The second in our SWOOP Video Blog Series:

Slide 1

Hi there, I’m Laurence Lock Lee, the co-founder and chief scientist at Swoop Analytics

In this second episode of Swoop Benchmarking insights we are drilling down to the Yammer Group level. Groups are where the real collaborative action happens.

As Yammer Groups can be started by anyone in the organisation, they quickly build up to hundreds, if not thousands in some organisations. Looking at activity levels alone we will see that the majority of groups do not sustain consistent activity, while a much smaller proportion look to be really thriving.

As useful as activity levels and membership size are, as we have suggested before, they are crude measures which can mask true relationship centred collaboration performance being achieved.

In this session we provide insights into how organisations can compare and benchmark their internal groups.

Slide 2

There is no shortage of literature and advice on how to build a successful on-line community or group. The universal advice for the first step is to identify the purpose. A well articulated purpose statement will identify what success would look like for this group or community.

What we do know from our experience to date is that there are a variety of purposes that online groups are formed. IBM has conducted a detailed analysis of their internal enterprise social networking system, looking to see if the usage logs could delineate the different types of groups being formed. What they found was five well delineated types of groups. {IBM classification from years of IBM experience  http://perer.org/papers/adamPerer-CHI2012.pdf }

The identified groups types were:

  1. Communities of Practice. CoPs are the centerpiece of knowledge sharing programs. Their purpose is to build capability in selected disciplines. They will usually be public groups. For example, a retail enterprise may form a CoP for all aspects of establishing and running a new retail outlet. The community would be used to share experiences on the way to converging to a suite of ‘best practices, that they would aim to implement across the organisation.
  2. Team/Process. This category covers task specific project teams or alternatively providing a shared space for a business process or function. In most cases these groups will be closed or private.
  3. Groups formed for sharing ideas and hopefully generating new value from innovations. It is best to think about such groups in two stages, being exploration and exploitation. The network needs to be large and diverse, to uncover the most opportunities. However, the exploit stage requires smaller, more focused teams to ensure a successful innovation
  4. The Expert / Help type group is what many of us see as the technical forums we might go to externally to get technical help. For novices, the answers are more than likely available in previously answered questions. In essence, they would be characterised by many questions posted, for a selected few to answer.
  5. Finally, the social (non-work) groups are sometimes frowned on; but in practice they are risk free places for staff to learn and experience online networking, so they do play an important part in the groups portfolio.

 Slide 3

This table summarizes the purposes and therefore value that can accrue from the different group types. Some important points that can be taken from this are:

  • Formally managed documents are important for some group types like CoPs and Teams, but less so for others, where archival search may be sufficient
  • Likewise with cohesive relationships, which are critical for teams say, but less so for Expert/Help groups for instance.
  • Large isn’t always good. For idea sharing the bigger and more diverse, the better. For teams, research has show that once we get past about 20 members, productivity decreases (https://www.getflow.com/blog/optimal-team-size-workplace-productivity)

 Slide 4

More than 80 years of academic research on performance of networks could be reduced to an argument between the value of Open and diverse networks versus closed, cohesive networks. This graphic was developed by Professor Ron Burt from the University of Chicago Business School, who is best known for his research on brokerage in open networks. However, Burt now concedes in his book on Brokerage and Closure in 2005, that value is maximised when diversity and closure are balanced.

It is therefore this framework that we are using for assessing and benchmarking Yammer Groups.

Slide 5

For pragmatic reasons we are using group size as a proxy for diversity, with the assumption that the larger the group, the more likely the more diverse the membership will be. For cohesion, we measure the average 2-way connections/member, using the assumption that if members have many reciprocated relationships inside the group, then the group is likely to be more cohesive.

This plot shows a typical pattern we find. The bubble size is based on group activities, so as you can see, activity is an important measure. But the positioning on the network performance chart can be quite differentiated by their respective diversity and cohesion measures.

The pattern shown is also consistent with what we see in our prior network survey results, which essentially shows that it is difficult not to see diversity and cohesion as a trade-off; so the ideal maximum performance in the top right corner, is in fact just that, an ideal.

Side 6

Now if we overlay what we see as ideal ‘goal states’ for the different types of groups that can be formed, it is possible to assess more accurately how a group is performing.

For example, a community of practice should have moderate to high cohesion and a group size commensurate with the ‘practice’ being developed.

The red region is showing where high performing teams would be located. High performing teams are differentiated by their levels of cohesion. Group size and even relative activity levels are poor indicators for a group formed as a team. If your group aims to be a shared ideas space, but you find yourself characterised as a strong team, then you are clearly in danger of “group think”.

Likewise you can infer a goal space for the Expert/Help group type.

If you are an ideas sharing group you have an extra measure of monitoring the number of exploitation teams that have been launched from ideas qualified in your group.

For the group leaders, who start in the bottom left, and many who are still there, it becomes an exercise in re-thinking your group type and purpose and then deciding the most appropriate actions for moving your group into the chosen goal space.

For some this may be growing broader participation, if you are expert help group; or building deeper relationships if you are a community of practice or functional team.

Slide 7

So in summing up:

Groups come in different shapes and sizes, where simple activity levels and membership size are insufficient for assessing success or otherwise.

Gaining critical mass for a group is important. Research has shown that critical mass needs to also include things like the diversity in the membership and the modes used to generate productive outputs.


The Diversity vs Cohesion network performance matrix provides a more sophisticated means for groups to assess their performance, than simple activity and membership level measures.

Once group leaders develop clarity around their form and purpose, the network performance framework can be used to provide them with more precise and actionable directions for success

Slide 8

We have now covered benchmarking externally at the Enterprise level and now internally at the group level.

Naturally the next level is to look and compare the members inside successful groups.

Thank you for your attention and we look forward to having you at our next episode.

Seeing How You Work, Changes How You Work – What’s Your Online Persona?

Our SWOOP Personas are having a much bigger impact than I expected. For a quick summary of the five personas see our previous posts: Observer, Broadcaster, Responder, Catalyst and Engager. In summary, these personas provide you with insights into your online behaviour on your enterprise social network.

I recently spoke to a community manager about this, and he told me this wonderful story about the impact the personas has had in his organisation. One of his colleagues, a senior manager, had been receiving help from a communications specialist to write updates on the Enterprise Social Network.

However, when the community manager showed the senior manager how little she had used the ‘Like’ feature, she realised two important things. Firstly, she was missing out on the positive recognition a ‘Like’ can provide the recipient, especially in her role as a senior manager. Secondly, she realised that she couldn’t outsource posting, replying and liking to her communication specialist. Interacting on an enterprise social network is a deeply personal thing, and as ex-CEO of Telstra David Thodey told us in a recent interview, he found the most important thing in generating transformational chance was to have authentic conversations with staff. The senior manager now does her on posting, replying and liking, and for me this really shows that:

Seeing how you work, changes how you work.

In CUA, an Australian bank piloting SWOOP to drive adoption of their Enterprise Social Network, they also saw the power of these simple personas in creating a common language so you can think about what you do, and what collaborative profile would be most effective for you. For instance, a communications specialist might operate best as a Broadcaster and a technical expert as a Responder. We generally consider the Engager to be the persona that all people managers would want to be, but a lot of the real value lies in reflecting over what you are, and what you ideally should be.

What is your SWOOP Persona?

By now you might be wondering what your own persona is. Answer the following questions to get started. Please note that your persona is not dependent on volume of your online activities, but the relative spread of what you do (post/reply/like) and what you get back.

When I am using my Enterprise Social Network… Not like me Some-times like me Like me
1. I post links to, or attach, interesting content I think people want to know about
2. I post updates to my team/colleagues about things they need to know
3. I call out colleagues for a job well done
4. I ask people for help with problems/challenges
5. I reply to requests for input/assistance where I can add value
6. I often prompt people to participate (@mention/notify others)
7. I hit ‘Like’ whenever I see a post/reply that I like, or something I want to show support for
8. My posts always get replies and/or likes
9. I am often encouraged by others to add input (am being @mentioned/notified)
10. I read posts, but don’t participate myself

Now, review your answers and determine which persona you think you are. Is that what you’d like to be?

SWOOP Personas

If you are with an organisation that has SWOOP running, then you should jump in and have a look at if your self-perception mirror reality. I’ve always thought of myself as an Engager, and must admit to you that I was pretty guttered when I saw that I was a Broadcaster on our network. My knee-jerk reaction was “Why aren’t you responding or liking the stuff I post!”, but my wonderful co-founder Laurie Lock Lee calmly said “Well – maybe you need to think about what you’re posting.”. I, of all people, should know this. I mean, we actually created the SWOOP persons to provoke this exact conversation, but it still hit me pretty hard as it was suddenly about what I was doing and not about what other people weren’t doing. It got very personal. I started to reflect over the posts, and replies that I had been making, and thinking about ways to make it more engaging. I tried to ask for more feedback by @ mentioning people, and also started to think more about what actually generates value for others rather than just focusing on things I think they need to know.

By seeing how I worked, I managed to change how I work. For the time being I am an Engager, but I know I’ve got to keep an eye on my persona to ensure that my changed behavior is locked in. This is not set and forget just yet!

Not on SWOOP yet? Try our 2 week free trial to check it out and get your SWOOP persona.

AirBnB vs Booking.com….your preference?


Having spent the last week immersed in ‘Platfirms’ (platform businesses) I now have some time to reflect as we leisurely wind our way through the French Pyrenees. In preparing for this trip I drew on my favourite accommodation platforms Airbnb and Booking.com. I’ll be up front and say I do have a preference for Airbnb. However, Booking.com tends to come into its own when looking for more affordable accommodation in the larger cities or towns. Putting this aside however, and reflecting on one of the key messages I have been making around relationship centred measures over simple activity measures, I can see a subtle but clear difference between the two platforms when it comes to relationships. Both sites are B2C businesses looking to facilitate strong relationships between their suppliers (properties) and customers (consumers) through the social media they enable. I recently published this diagram to identify potential relationship connections being facilitated by platform businesses.

Social Analytics

In this age of social, all of us are potential critics, and our comments form a major part of the buying decision for our fellow consumers. The relationship between ourselves and our fellow consumers is however only passing. We mostly care about the content. We don’t tend look to build a real relationship with those making the comments. While I don’t have any experience on the supplier side of these platforms, I suspect the property owners aren’t necessarily looking to build strong relationships with fellow property owners, other than to perhaps learn how to best represent their properties to prospective customers, by viewing competitive properties. In other words, for the most part, social media B2C platforms are really only building quite shallow relationships, centred around the media provided, with the express goal of closing transactions. Nothing wrong with that. In fact, this is exactly why the platform approach has proved so stunningly successful in this B2C marketplace.

Now let’s go back to my personal preference for AirBnB over Booking.com. I can talk about the variety of properties and situations; sometimes the attractive pricing, but I want to focus in this case on the relationship I can form with the property owners. For many of us, the biggest risk we take in booking accommodation in far flung, previously unseen places, is the unknown of things that might go wrong. That’s why we tend to look more closely at the negative comments a property might receive. What differentiates AirBnB from Booking.com in my experience is the way it encourages you to connect and converse with prospective hosts i.e. build a personal relationship. I tend to do this, if only to provide assurance that the pricing and availability is up to date. AirBnb measures host responsiveness to encourage these interactions. It also gives the host the opportunity to ‘personalise’ the ‘transaction’. And the more you converse before arriving, the more likely it is that your stay will feel more like staying with a new friend, than simply a property owner. It is these experiences I tend to remember and recommend to my friends. Of course you can use AirBnb, just like Booking.com, as a booking service; and I have done this too. Most of the times it works out fine. But certainly the times when things have not worked out so well are when we have relied solely on the ‘media messages’.

To be fair to Booking.com, it does target commercial properties, where your contact is more likely a staff member, than the owner. That said, it wouldn’t do any harm if commercial property owners encouraged their staff to converse online with prospective customers, as if they were the owners. And when this is encouraged and even facilitated by the platform, it is less likely that the supply side would feel exploited. In fact we can see now that Uber is acknowledging driver groups banding together in a union style arrangement. Perhaps the addition of relationship centred metrics may also be able to predict platform performance in the B2C world as well?

We would be interested in hearing your relationship experiences with ‘platfirms’ like Airbnb and Uber.

Hong KongOn a lighter note on this trip we had a long day stopover in Hong Kong, so we were looking for an inexpensive hotel to freshen up during the day. We did use Booking.com, which duly provided us with many options along with many comments about the compact sizes of the rooms and the mode of access, usually through a hawker’s market. All I can say is that if anyone were to come up with an innovation around ‘stand-up’ hotels, it will likely happen in Hong Kong!





The Age of the ‘Platfirm’

Platfirm? Is this a made up word? Well actually yes. It’s the term Open Knowledge have coined to describe businesses whose core business model is the platform. The ‘Platfirm Age’ is the theme of this year’s social business forum (#SBF16) held in Milan, Italy each year. Essentially a ‘platfirm’ is a firm that facilitates exchanges within a business ecosystem of suppliers and consumers. The popular examples of ebay, Airbnb, Uber and Amazon are regularly talked about in platform conversations. In opening the forum, Open Knowledge Co-founder Rosario Sica provided some compelling statistics to demonstrate that we are indeed in a new age of platform enabled business models. In Apple, Google, Amazon and Facebook we now have 4 of the top 5 companies in the world, by market capitalisation, being ‘Platfirms’. The issue of ‘Platfirm immigrants’ (as opposed to ‘platfirm natives’) was highlighted as a continuing challenge for traditional businesses wanting to get on the ‘from foot’, in either protecting or enhancing their businesses though platform thinking.

Platfirm blog

Keynote speaker Sangeet Paul Choudary was an excellent choice to lead the discussion. He and his co-author’s recent book on the ‘Platform Revolution’ provided the deepest coverage yet on just what makes platform businesses tick. Having read both this book and his first book Platform Scale I was pleased Sangeet was able to offer some new insights. One statement “What is good for the platform is bad for the ecosystem” brings to mind the power of a platform like Uber to be able to at times exploit both its suppliers (drivers) and customers with its pricing policies. One could appreciate now that powerful platforms relying purely on market forces could have the same negative impacts as a monopoly provider. On this point Sangeet was in favour of some regulation to guard against inappropriate use of market power. The other plea from Sangeet was his desire to see platform business opportunities in ‘public good’ applications, especially in developing countries. I think I would summarise Sangeet’s speech as ‘do good, not evil’, a refreshing perspective from a genuinely nice guy.

Sangeet was followed by several others developing ‘platfirms’, too much to report here (not the least because many of the speeches were in Italian, and my Italian is limited to food choices!). One interesting application worth reporting on and I think addresses Sangeet’s ‘do good, not evil’ maxim was from Michele Casucci, founder of Certilogo. Certilogo’s mission is to drive fake products out of the market. Now we have all probably bought fake Italian products in low cost market places before; well at least I have. But I was fully aware I was buying fake products at a very low price. And isn’t imitation the best form of flattery? More insidious though are the fake products sold at genuine prices. And on reflection it’s probably a good idea not to buy any fake product, as I suspect that those manufacturing even the cheap imitations are being exploited for their efforts. So I think Certilogo certainly meets the ‘do good, not evil’ maxim. How does it work? Certilogo crowdsources information from people who have purchased, or are about to purchase from expensive brands like Versace, Moschino, Paul & Shark etc. and then matches that information with supply chain data sourced from the brands that they support. In this way they can authenticate a product as genuine or fake in a matter of seconds. Like a classic platform, there is value exchanged on both sides. The brands get to find out where fake imitations of their product are being sold. The customers can find out whether they are buying, or have bought a genuine product or not. And perhaps by driving the producers of fake product out of the market, they are helping remove less scrupulous manufacturers, who are potentially exploiting workers in developing countries, to develop their goods.

The HBR Italia supplement published in support of the forum is certainly worth a read and is available here in English.