Description

Opinion

Subject

There is a high probability that machine-learning applications will begin to be used in the support of career guidance and counselling soon. Since machine learning is very good at identifying patterns, relevant technologies could be used in matching skill profiles to job profiles when supporting people in identifying relevant jobs. It is very unlikely now for AI to be used in many other areas of career guidance and counselling though, at the moment, since the technology is not yet capable of basic sensemaking. For example, ChatGPT fails at answering the question, where a girl was born, if the only available information is that her mother was in Munich at the date of her birth.
 
Sophisticated sensemaking is needed in career guidance and counselling though, e.g., to understand what resources and motivations people hold (something they often need help expressing). It is therefore very unlikely that AI technologies will replace career guidance and counselling in the future. Of course, as usual, there will be offers by technologists, who hope to offer AI-based services to people, e.g., a variant of ChatGPT that pretends to help people identify relevant vocations, training, or jobs. In addition to the problem that AI applications are currently unable to assist people in clarifying their own needs and strengths (i.e., what people are unaware of themselves), large language models like ChatGPT are also unfit for the job for other reasons though. ChatGPT is known to fabricate and falsify studies that support its claims, e.g., in "summarizing" the history of bear's colonization of the universe.
 
Machine-learning algorithms also adopt biases and stereotypes from the data that they have been trained with: It would not be surprising for an app like ChatGPT to suggest that men ought to study engineering whilst suggesting that a woman with similar talents and interests ought to become a teacher. Finally, machine-learning algorithms lack a moral compass and cannot make sense of values. While machine-ethics specialists are trying to program relevant solutions, we cannot expect to find a "general moral intelligence" in machines any time soon. The problem here is that AI applications may cross diverse ethical boundaries in trying to "council" people. For instance, it is foreseeable that they could recommend exploitative work.
 

Benefits and challenges of the use of artificial intelligence in career guidance

AI applications are extremely good at making sense of extremely large volumes of data in a relatively short time. An area that can surely benefit from AI applications is labor-market research. Via machine learning, millions of published job profiles could be analyzed to identify short- and long-term developments and understand how these changes correlate with other trends. Labor-market researchers will be able to provide career professionals with much better information this way.

Another area where artificial intelligence is likely to improve career guidance and counselling is career education. AI is already used to individualize learning pathways in educational programs (e.g., language-learning software) and digital games. It can be used to guide learners through educational programs in ways through which they concentrate on relevant exercises, for instance, instead of learning about things that they know. Finally, AI applications may be used to improve the provision of career guidance and counselling in the near future, e.g., by identifying people with career development needs based on their internet-search behavior and providing them with targeted advertisements for career services that are likely to be of use for them (and not primarily for those offering the services).
 

How ensure that guidance experts and practitioners could be trained to effectively use AI in their practices?

In the near future, career professionals will be drawing on diverse resources in their practice, which are based or powered by machine learning. To prepare them to work with these resources, they will require what is often called a critical AI literacy. One of the most important lessons to be learned is that AI applications are not infallible and are prone to reproduce biased information or even create misinformation themselves. For example, apps like ChatGPT can synthesize information on vocations, e.g., required skills, typical activities, etc. However, there is a high risk that these apps could create false representations, based on erroneous data or misinformation that they collect from undisclosed sources across the world wide web. In the USA, judges have been overly confident in using AI-generated risk assessments that people convicted of crimes may engage in further crimes.
 
In consequence, people from poor communities have often been punished more harshly than people from more wealthy, predominantly White communities – solely based on the information that more crimes happen in certain zip-code areas than others. Career professionals working with AI applications will need to understand these kinds of risks and recognize red flags, i.e., signs that information or decisions from AI applications could be p