An article was released on the 26th of July on ABC with the title ‘Behavioural science quietly shaped Robodebt's most devilish details — and their work in government continues’. The article, as the title quite severely hints at, dives into how behavioural science was used to rob people blind through a scheme of ‘tax repayments’ as to take money away from citizens to pay the government, and the mess that ensued from this. At least that’s what it claims.
For context, as I assume not everyone reading is invested in Australian economics, let’s start with what the Robodebt fiasco is.
In July 2016 the Australian Government and its respective tax authority (ATO) released a new method of debt assessment (e.g. how much you owe the government (or they owe you) as part of being on a form of governmental benefits). Initially under or overpayments were calculated manually, but the model was shifted to becoming automated, using income-averaging. That’s where the ‘robo’ in Robodebt comes from. The issuing of debt notices (meaning you owe the government money) was also automated. This was based on an automated data-matching system that compared Centrelink (Australian Government Services Agency) records with averaged income data from the ATO.
You should be able to spot the fog through the trees. This system was flawed beyond belief. The algorithm wasn’t right, an unimaginable number of people were informed they owed the government thousands of dollars (which is an amount of money most people on benefits simply do not have) and the scheme was ruled as ‘unlawful’ in 2020, having to refund at least 470,000 unlawful debts, to the tune of 721 million (!!!) Australian dollars.
Now you might be wondering: ‘where’s the behavioural science in all this?’ Because behavioural science doesn’t really tend to involve itself with algorithm creation and validation.
Well in the article released by Peter Martin he reckons that the reason Robodebt went ‘tits up’ (my words, not his) had everything to do with the design of the letters people were sent to inform them of their thousands of dollars of debt (no, it wasn’t the algorithms fault at all…). He claims, and he mentions ‘evidence from the Royal Commissioner (Catherine Holmes) tasked to look into what happened with Robodebt, that it was ‘behavioural science experts’ that pressed several of their ‘insights’ onto Centrelink for the design of the letters. And the kicker here: it was ‘behavioural experts’ who urged Centrelink to remove the phone numbers from the letters, as to have people ‘quietly pay up, or go online and provide years of payslips they probably didn't have, rather than jam up its switchboards asking questions’ (his words, not mine). Issue is, this line of reasoning makes very little sense. And it doesn’t make sense for a variety of reasons:
No behavioural scientist is going to recommend to have contact information removed. That’s key information to support people in their decision-making process, especially when they do not possess high technological or online literacy. This is not a nudge, it’s a sludge. Which brings me to:
The only governmental ‘nudge unit’ who’s likely to have been affiliated with this project would be BETA – who do get mentioned in this article. BETA are behavioural scientists FOR GOOD. They are very unlikely suggest this.
I have spoken to people from BETA and they had nothing to do with the entire project. Not only did they never recommend this specific sludge, they never worked on the project. Of course, I wouldn’t just take people’s word for it, but I also genuinely cannot find proof of their involvement anywhere online.
The fact that neither Peter Martin, nor the commissioner report states which behavioural experts (i.e. individual names, units, consultancy firms) worked on this project makes me wonder if they even exist and whether it wasn’t someone internally who read Nudge once upon a time.
And even if there were behavioural insights recruited for this project as per the ‘email evidence’, it’s never specified what the insights or the suggestions were. Martin’s article itself mentions that the ‘behavioural experts’ recruited also recommended that Centrelink sent out colour coded letters. But that never happened either as all was sent out in black and white. Which leads me to the last point:
Who on earth signed off on this? A ‘behavioural recommendation’ is one thing, but someone had to approve it. There are standard parts to templates for communication and the pieces of information they are required to include, by law. And although I’m no expert of Australian law, I feel like a phone number to call is one of those. So there should have been a communications, strategy, risk or PR unit (or whatever the government equivalent is) to sign off on this. It should never have passed through.
Reading more objective and better-informed articles on the topic reveals that Royal Commissioner Catherine Howard has nothing positive to say about the Australian Government’s attitude and procedures with regards to this scheme. She is being quoted in the AFR as calling the government ‘soulless’ . And that seems to be a fair assessment. No mention of behavioural science though.
Now, back to the behavioural science of it all and the article written by Martin. The entire article is anti-government, and that way that quotes and insights are used out of context is baffling. Behavioural science is even being criticized for trying to change the colours of the letters. But that’s just behavioural design.
Anyway, that’s not the problem. My problem is the way the article is written, because it’s written very cleverly, but the argument isn’t sound. It first mentions that Robodebt was awful (it was), then that behavioural insights were sought (needs to be substantiated) and that those insights specifically lead to the removal of the phone number (evidence presented doesn’t show that). There’s nothing there. It’s fluff. Angry fluff.
The article then moves onto the powerlessness of the people who were impacted. The lives that were ruined. Those who took their own lives. From anger we now move to sorrow. Guess what? The article is written behaviourally – you're meant to feel the article, not actually study it...
Now we dive into BETA (Behavioural Economics Team Australia). The unit that was created in 2015 to help the AU government. It was modelled after the UK’s nudge unit (Behavioural Insights Team, 2010) which also started in government. Then there’s a mention of choice architecture (even Cass gets thrown in there for good measure). Why? No reason – but you need a specific entity to blame. A scapegoat needs a face.
And that’s where we hit the grand finale. Is it truly an epic saga of behavioural science trying to destroy the world without there being a mention of the replication crisis and Ariely’s (now retracted) work? Of course not (almost ironically, Martin does like psychology, which is where the replication crisis started, but who cares about actual facts in this epic tale of doom). It ends with the announcement of the Australian Centre for Evaluation (ACE), for which Martin also has absolutely no hope.
Cool.
Ultimately, this is a poorly researched article, which mentions BETA in a way where it seems to accuse them, but tries really hard not to do that explicitly (because then you can sue for libel). Martin just doesn’t like behavioural science. He calls the science hollow, without an interest in how interventions impact people, without empathy. Which I find a joke coming from someone who’s trained as a neoclassical economist, who’s Twitter feed tells me he’s very anti-government. But the latter statement is my opinion – and I wouldn’t dare call it journalism.
Quick conclusion: behavioural science had little if anything at all to do with the Robodebt fiasco – as from the evidence we have been presented with so far. Our field is going through a hard time. We know this. We’re not proud of it. We are doing our best to weed out the bad apples and sometimes, start from scratch again. But behavioural science can have really positive impacts on people’s lives, because we do try to understand people – their motivations, their barriers and design around those, and then evaluate how (if!) it worked and why (not). That’s the whole point of the science. And I will not have it libelled by someone who refuses to do something as basic as research into what actually happened.
Which leads me to my final note: people took their own lives because of the Robodebt scheme. Because they thought they were thousands of dollars in debt. Because they couldn’t cope. And Martin’s right – we should be emphatic towards that. We should give a damn. So don’t put it in a 400-word fluff piece that’s very much just a crusade against behavioural science. Show some respect.
