AI & Visa Applications

There is no doubt that artificial intelligence or AI is here to stay and in fact, it is slowly finding its way in to almost every aspect of our lives. Washing machines that can work out to wash you clothes, cars that (almost) drive themselves and applications that you can talk to in case you might get a little lonely. The rapid rise of AI and its integration into our daily lives is pretty impressive and also potentially a little bit scary.

Whilst these systems have caught on very quickly, it is also important to remember that they are just systems and despite our quick adoption of the words artificial intelligence, we actually haven’t achieved that yet. What we have achieved is enormously sophisticated programs that can rapidly compare, contrast and interpret vast sets of data, to give you what appears to be an intelligent answer.

There is no doubt that these systems are gradually making their way in to the Visa process and in fact they have been for many years. Even before the most recent system (ADEPT) was built, AI and the use of automated visa processing had been introduced. The rise of systems like ChatGPT and its various offshoots and competitors has meant that the available technology is far more complex and sophisticated.

Whilst some elements of these systems are included in our visa processing systems, applications are still largely handled by human beings. Some aspects such as identity checks, the progress of an application through INZ’s various systems and certain risk profiling criteria are handled by computers, but people still remain at the pointy end of the process.

However, there is every chance (and certainly a desire) by most Governments to utilise AI to take over significantly more the visa process than is currently happening. After all, it has the potential to be more efficient and at a far lower cost. However is it possible and what might that mean for future applicants?

AI Visa Processing

In terms of AI being used in the visa process, it is important to understand that in the current environment, the actual use of these systems is relatively limited. We don’t yet have ChatGPT deciding who does and does not get a Visa. However we do have the foundations for computers, programs and automated systems to be involved. Section 29 of the Immigration Act 2009, outlines that an automated system may be used by INZ to rank an expression of interest, process, grant or refuse to grant an invitation to apply, process an application for, grant or refuse to grant a visa and them same for entry permission.

Automated Visa Processes

INZ does use some automated processing already but it is very limited and also very basic - that however could change.

These permissions are then further entrenched in the operations manual, which is essentially the handbook to processing visas for INZ and applicants (although most applicants don’t really know the operations manual exists).

The key to this piece of the Act and also our instructions is that it allows for an automated system (which is very broad term) to effectively make a complete decision on an application or any aspect of an application. If you think about it, that is a pretty wide-ranging set of rules, and enables INZ to deploy any number or range of systems in the decision making process.

So far, the extent to which this has been used is limited. In fact in most cases the automated system is literally just a process of checks against information on file, which determines the outcome. A good example is the “interim visa” process which is largely automated. If you are here on a temporary visa, due to expire and have lodged a new application for another temporary visa, the system looks at your file and assuming there are no issues, sends out your interim visa. If there are issues, it goes to a human to decide.

Because applicants are unique and often present complex situations there is no real system (at the moment at least) that can manage that sort of assessment and so instead people remain the main processing engine. However, imagine if in the future, INZ decided that instead of Bob from the Auckland branch considering your application, ChatGPT would be given that task. It’s not impossible and whilst it probably wouldn’t be ChatGPT, it would be using essentially the same sort of processing system.

Now imagine if that system declined your application and you wanted to challenge that decision - except who are you are appealing to exactly?

INZ has been challenged on this before, scrapping a tool that assessed and prioritised deportation proceedings, back in 2018. The system used data models compiled by INZ to establish risk and harm and effectively who to take compliance action against first. It was removed due to accusations that the models were biased towards a particular demographic. This stuff is never as easy as the tech promoters would have you believe.

Visa Applicants Using AI

On the other side of the equation are the applicants, of which some are using AI tools to deal with INZ. A bit like cheating on that essay question, some applicants who are faced with complicated and difficult letters from INZ, outlining numerous concerns, AI has become a well used tool to respond.

AI Responses to INZ

There is a growing trend amongst applicants to have AI assist them in responding to INZ concerns, a bit like cheating on your homework.

I can understand why, particularly given many of INZ’s letters are written in a way that would make Stephen Hawking scratch his head. Most applicants have no idea what the actual rules are for their application and yet INZ bombards with them copy and paste sections of those rules.

So trying to understand where you went wrong, whether INZ’s concerns are valid and then how to respond to any of it, can be incredibly confusing; and when so much is at stake, it would be very tempting and often very easy to reach out for some AI assistance.

The problem is, AI doesn’t always work this out properly either and I have seen a fair few responses sent to INZ, that sound very eloquent, if not a bit robotic, but have in fact completely missed the mark. INZ are also aware of this happening and have been known to raise the AI response to concerns as a concern in itself.

Like most things that involve human beings, and the visa process is all about people, having AI trying to argue your case, isn’t really a smart move. AI might make your response sound better, but it isn’t going to be able to work out all the specific details you might need to provide, or whether INZ’s concerns are right in the first place. There is also a very valid question as to whether this is acceptable, given you aren’t the one responding.

AI can of course help you to polish up your own response, particularly if English is not your first language, or it can be used to interpret INZ’s letters for the same reason, but arguably the best way to respond to these sorts of concerns is still with the use of good, old-fashioned brain power and experience.

The Future Of Visa Applications

I have no doubt that over time, and as AI develops, its role in the visa process will increase. The foundations are there and the rules and the Act already make provisions for the use of automated systems. The key question will be in how far these systems go in terms of trying to assess applications and more importantly applicants.

In some areas of the visa process, an automated system makes a lot of sense. It removes unnecessary double-handling of certain tasks, removes human error and delivers a much greater level of consistency. It should also be a lot quicker and that then leads to faster visa processing times.

However there are other areas, where AI could become incredibly problematic. For example, imagine if your application was processed up to the point where INZ need to consider a medical waiver. In this scenario, the benefits of you being approved, which covers a very wide range of factors are assessed against the potential negatives (cost, burden to the health system and so forth). This is a very subjective assessment and there is no “one size fits all approach”. Will an AI model lean too far in one direction than the other, can it be balanced and objective, can it consider what is fair and reasonable. Arguably there are plenty of immigration officers who might struggle to achieve the same outcome, but I would put my money on a human being over AI any day of the week.

We obviously haven’t seen real AI turn up yet either and I am of the personal view that might not happen, but the systems we have now are developing at such a rapid speed that their use in processes like immigration are inevitable. The next time you call INZ’s call centre, ask yourself whether that voice on the other end of the phone is real or perhaps not…

Until next week!