Algorithms/AI in recruitment and non-publication orders in employment proceedings
In Spiga, a full (four-judge) bench of the Employment Court considered the correct approach to the power of the Employment Relations Authority and the Court to order that “all or any part of any evidence given or pleadings filed or the name of any party or witness or other person not be published” (Employment Relations Act 2000, sch 2 cl 10 and sch 3 cl 12). All four judges agreed that a permanent non-publication order should be made in respect of the plaintiff’s name and identifying details in this particular case; however, the judges were split as to what was the correct approach in reaching that result. A three-judge majority considered that the general approach to non-publication in the ordinary civil courts should apply in both the Authority and the Employment Court; whereas the fourth judge would have recalibrated the approach in the specialist employment jurisdiction based on the scheme and purpose of the Employment Relations Act 2000 and how that informs the broad discretionary power available.
One aspect of both the majority and minority judgments was the potential effect that publication of the name of an employee party (or even witness) in employment proceedings could have on future job prospects. Various issues were addressed on this point, one being the potential “blacklisting” of current/prospective employees through increased data analysis of employment judgments.
The majority view was that the evidence did not support a conclusion that this was currently a practice of significance in New Zealand, although that might change as the use of statistical and machine-learned algorithms increased:
“[73] … evidence was filed by the Employers and Manufacturers Association (Northern) Inc (EMA) that showed that internet searches of decisions were undertaken by some employers, and that some of them regarded information obtained as relevant or essential. Such a practice is not inconsistent with employers looking at substantive findings of the Authority or the Court. Employers may well have a legitimate interest in knowing relevant found facts about potential employees contained in decisions of the Authority and of the Court. These decisions, written by objective decision-makers in a judicial process, may be seen as more reliable than references, which tend to be from referees selected by the candidate who the candidate assumes will provide positive comment.
“[74] We accept also that there are cases where there is express evidence to show that publication has impeded a party’s ability to obtain new employment. But again the evidence does not extend to the reasons employers may have been cautious about employing the people concerned. There is some suggestion that just being named in a decision (for example, as a witness) can result in a form of ‘blacklisting’, but we have seen no evidence to support a conclusion that this is a current practice of any significance in New Zealand.
“[75] We have not seen evidence to enable us to conclude that there is an increased ‘chilling effect’ – that employees are increasingly not pursuing claims out of fear that they will be ‘blacklisted’. As noted in the submissions from the EMA, the number of proceedings in the employment jurisdiction has not reduced since the introduction of more widespread internet availability of decisions and social media discussion.
“[76] This position may, however, change. Research has indicated there is a danger in online publication being used by statistical and machine learned algorithms. These are increasingly being used by employers to make hiring decisions. Researchers from the United Kingdom state that these algorithms work by drawing on data about a given population with a view to identifying correlations between certain characteristics, and combinations of characteristics, and particular forms of behaviour. Thus, predictions are generated about how likely an individual is, given their characteristics, to engage in a specific behaviour in the future. If details about claimant characteristics are made available to the public, then there is scope for employers to rely on such techniques to predict how likely an individual is to take them to court (to challenge unlawful decisions and/or engage in litigation in the future) with a view to refusing them employment on that basis.” (footnotes omitted)
In contrast, the minority view was that (in terms of the Authority/Court’s broad discretionary powers) there is already sufficient evidence of (or at least information about) recruitment practices – including elements of “blacklisting” – to draw an inference as to the likelihood of damage to future job prospects from publication of the names of employees and witnesses (at [160]–[165] and [170]–[172]). As to the particular relevance of technology and the pace of change, the judge said:
“[180] There have been significant changes that have occurred in society and technology since earlier judicial observations about the appropriate approach to non-publication of names in the employment jurisdiction. One example is the emergence of artificial intelligence (AI) in the employment context to assist in making management decisions, such as recruitment.
“[181] While Business New Zealand (tenth intervener) submitted that its members (and more generally employers in New Zealand) do not have access to, nor the desire to access, these types of AI resources, the technological potential, when viewed alongside the evidence that it is relatively common for employers to utilise previous employment dispute records to screen job applicants, suggests that it would be timely to grapple with this issue now.
“[182] Developments overseas reinforce the point. By way of example, one group of leading academics observe that machine learning algorithms may learn from patterns in previous employment disputes and use that to inform hiring decisions (which could lead to bias being baked into or exacerbated by automated decisions). A concern about a developing ‘justice gap’ for workers, compounded by the lack of transparency characterising most automated or semi-automated decision-making processes, has increased the information asymmetries in the already imbalanced employment relationship. A related concern is that algorithmic opacity may conceal the violation of the rights of those workers subject to algorithmic tools (including reducing the likelihood of discrimination being perceived). As [counsel for the Privacy Commissioner] submitted, due to the significant advances in AI and the expanding breadth of information readily available online, the potential for its use (and misuse) is likely to also expand.” (footnotes omitted)
Amidst the current purple patch in AI developments, it is interesting (to the writer at least) to note this particular aspect of Spiga. Footnotes in both the majority and minority judgments point to further reading on the topic, including Colin Gavaghan, Alistair Knott and James Maclaurin The Impact of Artificial Intelligence on Jobs and Work in New Zealand (University of Otago, 2021), a report produced as part of a three-year project on AI and law funded by the New Zealand Law Foundation (see Artificial Intelligence and Law in New Zealand <www.cs.otago.ac.nz>).
An account of the actual hearing in Spiga is also of interest: see Reweti Kohere “Employment Court hears claims of employee blacklists in test case on name suppression” (13 October 2023) The Law Association <thelawassociation.nz>. So too are earlier opinion pieces from lawyers on non-publication orders in employment proceedings, such as:
- James Crichton “Advocating for change – name publication in the Employment Relations Authority” (17 November 2020) Three60 Consult <three60consult.co.nz>.
- Susan Hornsby-Geluk “Should court give name suppression in employment cases?” (24 March 2021) Stuff <www.stuff.co.nz>.
- Alastair Espie “Why workers should have the right to remain private” (28 September 2021) BusinessDesk <businessdesk.co.nz>.
- David Burton “Is it fair to name workers who take their boss to employment court?” (1 June 2022) Stuff <www.stuff.co.nz>.
A proposed Member’s Bill in the name of Labour MP Helen White and drafted (it seems) in 2022 may also be noted: see New Zealand Parliament “Employment Relations (Privacy of Parties to Proceedings) Amendment Bill” <bills.parliament.nz>. The explanatory note states (in part):
“This [Bill] will encourage all those involved [in employment proceedings] to participate without fear of publicity after the resolution of the matter. It recognises that resolving employment relationship problems has both a public and a private aspect and strikes a balance between these interests. It recognises a changing world where the ability of third parties to access the names of people involved in employment cases is much greater and can harm participants. The public will still be able to learn from any precedent value of examination of the application of the law to similar circumstances. The names will only be published where this is valuable to the public.”
(This Bill is merely proposed at present; it has not been selected from the ballot for consideration by Parliament.)
At any rate, the majority judgment in Spiga is now the leading decision on non-publication orders in New Zealand’s employment jurisdiction. For commentary on Spiga in Westlaw New Zealand, see Employment Law (online ed, Thomson Reuters) at [ER189.04A], [ERSch2.10.01] and [ERSch3.12.01].