Running the gamut of emotions on a scale of 1 to 5?

The title, given that this post is written in response to an external blog post (Matthew Hammell’s very thought-provoking “Sentiment isn’t just for Sympathy Cards!”), is somewhat cheeky, but I hope that I can go on to demonstrate that an IT-based response to the ‘softness’ of HR issues can – and perhaps inevitably will – trigger an HR-related reaction that can’t help but wonder about the ‘hardness’ of IT solutions.

Matthew, on the basis of his profile, has a long and successful track record in the corporate sphere in both HR and IT: he knows, we can safely assume, whereof he speaks. As he acknowledges, there are really two issues at play in the area he is currently addressing. The superficial problem – to use the wrong adjective, I admit, but to call it ‘the presenting problem’ would be to potentially confuse management speak with psychologists’ jargon – is how to gauge opinion, feeling and atmosphere amongst the human resources (or ‘people’, as we refer to them outside the office). But there is an underlying problem: the opinions and thoughts that most need to be swayed are those of the occupants of the C Suite, for whom measurement is a matter of firmness, definitiveness and bottom lines. They are not a group of individuals much moved by ‘data’ such as “We have had some informal feedback to the effect that …”

Matthew is exploring the use of ‘sentiment analysis tools’: software that mines emails and other digital sources, helping OD professionals to pick through potential mountains of ‘data’ to recognise patterns of assent or dissent (or, perhaps worse, indifference or prevarication). The idea is certainly an interesting and intriguing one given his starting point: the frequent failure of merger and acquisition (M&A) activity to deliver on its potential promise, as ‘the same boat’ that everyone theoretically winds up sharing so often then runs aground on the metaphorical rocks of human difference and disagreement.

The tendency towards M&A failures and the reasons for them are something we’ve previously explored in this blog: Chris Rogers wrote on the topic in 2009, identifying seven main obstacles to success:

  • impact on both businesses’ ability to sustain financial performance
  • loss of productivity and distractions to business drivers as the merger  is rolled out
  • a clash of cultures that drives negative behaviours and business loss
  • push and pull factors that can lead to the loss of top talent
  • positioning for power and a clash of management styles and/or egos
  • lack of management understanding of the different dimensions of change and  how to enact change management
  • failure to appreciate the need to synergise the changes of process and people
  • poor understanding and/or communication of objectives.”

While this list may be different from Matthew’s, our conclusions – that ‘soft’ issues and their skillful management lie at the heart of the matter – are in alignment. I confess, however, that I start to drift out of alignment with Matthew as I consider the application of technology to ‘solve the problem’. Consider, for example, Matthew’s following explanation of ‘sentiment analysis’:

It’s a software driven process that analyses text and identifies feelings, reflections, likes and dislikes. At its most sophisticated it can give you a very good temperature reading and a good sense of what peoples attitudes are to particular issues. It provides you with an understanding of the judgements people are making, their inclinations, passions and opinions.”

Like Matthew, my background straddles both ‘soft’ areas and IT: I have worked in digital publishing, web and software development and managing the introduction of large-scale workflow automation systems. Automating, for example, the admissions system of a large, multi-site University, I could (had the tools been available at the time) run an analysis of the emails I received from academics and administrators across all the various Schools, Faculties and campuses. It might – or might not – have been enlightening: I have qualms, as someone who works primarily with language, about the ability of textual analysis to identify subtleties of human communication that include sarcasm and irony. (Our occasional misreading of emails, deprived as we are of non-verbal communication signals, should perhaps alert to the fact that software analysis ‘reads’ with only the emotions that it has been programmed to detect – the contextual basis is even smaller.)

But I think there are some different underlying issues – and they are for HR, managers and leaders. One of these is that ‘digitising the carpet’ – thinking of text as something electronic, although that may be a large part of the reality of internal communication in many organisations – can also digitise our tendency to sweep things under it. One of the reasons that inaccurate assumptions (including, of course, ‘the merger will go smoothly’) go unchallenged is that opinions go unsurfaced, or that varying points of view are not given opportunities to be voiced. Am I the only person sufficiently paranoid to think that the adoption of ‘sentiment analysis’ might, in itself, be somewhat inhibiting: “your emails and intranet postings will be analysed to identify your feelings and emotions” is a sentence that might make quite a number of people feel inclined to step away from the keyboard. (On the subject of online monitoring, paranoia and dissent, you might wish to have a look at our review of Evgeny Morozov’s The Net Delusion.)

Matthew also makes passing reference to the likes of Trip Advisor, which raises another concern in my mind. Trip Advisor’s openness to ‘spoiler’ comments from rivals isn’t quite it – it’s more that it offers the flipside of the ‘Facebook ‘Like’’ coin that Morozov identified. While Facebook offers us the ability to ‘like’ something – without nuance – at a single-mouse click, other screen-based phenomenon make it equally easy to pour scorn or grievance on something. And I can’t help but think that passionate disapproval is more likely to tempt us to our keyboards than a mild but not especially pressing sense of approval. (The word ‘rant’ is, after all, hardly ever applied to an outpouring of praise.)

A third concern is, appropriately enough, somewhat more shadowy: lurking – the phenomenon of people reading online forums or social media but who never actively participate. From twenty years’ experience of online communities of all kinds, it’s a truism that most are dominating by the voices of only a small percentage of those either with access or actively reading. Although the rise of social media may lead us to think that the entire planet is now incessantly babbling, the distribution curve of babble-per-person is worth remembering. Despite the advent of Twitter, the majority of us have yet to express our opinions on any number of things. Perhaps this is a blessing, but any ‘sentiment analysis’ we might wish to undertake might be heavily skewed if this is not taken into account. (If an analogy helps, consider Yasman Alibhai Brown in today’s Independent: the much commented 24% of votes cast for UKIP in recent UK local elections is actually 24% of 32% – the percentage of those eligible who voted. 92% of the electorate did not cast a vote for UKIP, yet the party’s policy platform currently dominates political discussion. How accurately would a sentiment analysis of current online political discussion represent the spread of opinion across the whole electorate?)

To resurrect an ancient cliché, our tools have a tendency to shape us and our actions: when the purpose of a tool is to inform us, we should not be afraid to question the sources that a tool has drawn from in delivering its answers. Nor should we be afraid to remind ourselves that it is preferable that we use the tools rather than the other way around (and our tendency to be beguiled by technological breakthroughs should sound another note of caution).

And yet … all this is not to pour nothing but cold water on Matthew’s idea. As he has clearly said, the main causes of disappointing results in M&A activity are down to ignoring human factors. A tool that encourages HR – if no-one else – to gauge opinion and to recognise the importance of doing so is a potentially useful addition to the arsenal (with the caveat that applies to most findings delivered by unseen algorithms – use them as levers for better questions rather than taking them as answers). But I think the rest of the HR quiver needs review here, and I offer them some further questions:

  • Are you harvesting free-form ‘opinion’ or are you posing actual questions (a questionnaire gathers more structured responses)
  • Are you checking the percentage and spread of the ‘electorate’ who are participating
  • Are you guaranteeing anonymity of ‘sentiment analysis’ (there’s little point seeking to canvas honest opinion if there’s even a perceived threat of paying a price for expressing it)
  • Are you talking regularly to line managers, and encouraging them to openly monitor the prevailing atmosphere and ‘morale’? Are they doing so? (Research suggests that the best leaders do so as a matter of course, rather than in response to a reminder from HR, but the latter is a start …)
  • Are you automating something that you are neglecting to do ‘offline’ – the Internet may be referred to as its updated version, but it is not quite the water-cooler (and perhaps more an addition to it than a replacement for it)

But I think HR needs to ponder a bigger question? How can you change the contribution that you make to your organisation so that your opinion is valued to the extent that it is taken in board when it is expressed in anything other than ‘hard figures’? And how (often) do you analyse the sentiments of the C Suite – and then act on your findings?

Link to original post

Leave a Reply