Ethics |

Uncontrollable GOLLEM-AI and the total decoding and synthesizing of reality

Sat, 15 Apr 2023 08:05:11 +0100

You have read that I am criticizing the discourse on AI here in the last weeks and I have made a call for regulation of AI in education. But I have really underestimated the potential harmful effects which are coming through the uncontrolled release of Generative Large Language Multi-Modal Models (GOLLEMS).

Tristan Harris and Aza Raskin of the are developing in the video below a very well-grounded summary of the catastrophic risks of the current generation of AI on many fundamental aspects of societies and they are right to demand responsibility and a discourse about potential unwanted effects of large powerful tech-companies striving for market domination through a rat-race for the most powerful GOLLEMS.

If you have one hour today, on the dangers of the current generation of AI. Watch it and talk about it!

Why AI-based tools are never neutral

Sat, 28 Jan 2023 04:23:30 +0200

„ChatGPT itself is a neutral tool, and how it is used depends on the intentions of those who use it.“ (Cohen, 2023).

This messages is to its 125 000 members (estimaton) regarding an AI-based tool which is currently pushing the discourse on AI in education. While the one camp is calling out the next revolution for education, it was predictable that the other camp will soon say that technology has no effect itself on education. Both positions are signs of a technological determinism (Chandler, 1995) and both are dangerous as take-home messages for educators. I wish every educator would deal with the phenomenon of tecnological determinism to avoid these easy conclusions.

While these standpoints between the extremes “revolution” and “complete neutrality” are problematic for all learning technologies, for AI-based tools those are especially problematic since their reliance on datasets and algorithms leads to many ethical issues arising before we can even think about implementing them in an educational context. Leslie et al. (2021) speak about “cascading effects” when we take existing inequalities in the world as a starting point to build potentially discriminatory data which lead to biased designs and an increase of existing injustice.

Cascading effects of health inequality and discrimination manifest in the design and use of artificial intelligence (AI) systems (Leslie et al., 2021)

The Independent High-Level Expert Group on Artificial Intelligence (European Commission, 2019) discusses 7 interrelated dimensions which contribute to the establishment of a trustworthy and ethical AI.

Interrelationship of the seven requirements: all are of equal importance, support each other, and should be implemented and evaluated throughout the AI system’s lifecycle (European Commission, 2019)

These ethical concerns have led the European Commission to a proposal to qualify the use of AI in education as a high-risk usage scenario whith implications for their implementation.

„AI systems used in education or vocational training, notably for determining access or assigning persons to educational and vocational training institutions or to evaluate persons on tests as part of or as a precondition for their education should be considered high-risk, since they may determine the educational and professional course of a person’s life and therefore affect their ability to secure their livelihood. When improperly designed and used, such systems may violate the right to education and training as well as the right not to be discriminated against and perpetuate historical patterns of discrimination.“ (European Commission, 2021).

EU Artificial Intelligence Act: Risk levels (Telefonica, 2022)

I hope this makes clear what is at stake if we introduce AI in education and that the ethical dimensions are not comparable to for example those of introducing a learning management system. All in all, technology is never neutral and educators need the knowledge and skills to assess potential risks and ethical implications. For this purpose, existing frameworks for the development and implementation of AI-based tools can play a role (Vakkuri et al., 2021).

References

Chandler, D. (1995). Technological determinis and media determinism. Available under ( )

European Commission (2019). Ethics Guidelines for trustworthy AI. High-Level Expert Group on Artificial Intelligence. doi:10.2759/346720

European Commission (2021). Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL LAYING DOWN HARMONISED RULES ON ARTIFICIAL INTELLIGENCE (ARTIFICIAL INTELLIGENCE ACT) AND AMENDING CERTAIN UNION LEGISLATIVE ACTS.

Leslie, D., Mazumder, A., Peppin, A., Wolters, M. K., & Hagerty, A. (2021). Does “AI” stand for augmenting inequality in the era of covid-19 healthcare?. bmj, 372. doi:

Vakkuri, V., Kemell, K. K., Jantunen, M., Halme, E., & Abrahamsson, P. (2021). ECCOLA—A method for implementing ethically aligned AI systems. Journal of Systems and Software, 182, 111067.

This is not a post about ChatGPT but agenda setting

Sat, 21 Jan 2023 06:23:30 +0200

Since I am currently often asked about my opinion about ChatGPT I want to share why I do not write about ChatGPT.

The discussion about ChatGPT (Generative Pretrained Transformer) is at least as annoying as the Wordle wave that has hit my social media channels some years ago. Although we can put both examples under the category of “language games” (maybe in a different sense as Wittgenstein has described them) we can learn some more lessons about the public and partially scientific discourse of ChatGPT compared to the word game.

First of all, a lot of what I read are superficial reflections which seem to forget that AI in education is not a novelty. If you want to learn about the history of these kind of systems I can recommend by Mike Sharples (Sharples, 2022) and bis book “Story Machines” (Sharples & y Pérez, 2022). If you want to announce the next revolution, read this first.
Second, it would be good if users are becoming aware that they are becoming potential co-designers by their input and that feedback on the quality of answers or serious shortcomings is helping to improve the model and in the end the product which will most probably be commercial. It is by the way a special flavour that people are advocating rigorously for open and free learning material and that OER is the only solution for education and do not see the paradox that they are contributing to the promotion of a prototype of what will most probably become a commercial product and competitor for the most used search engine (Microsoft has announced to ). It is of course absolutely valid if researchers and teachers are “going with the flow” and show their involvement in the so called “hot topics” in educational technology. The problem is, that this discourse is pushing away the real hot topics in education: Access to education, quality of education and inclusion of non-standard learners into educational systems. The problem with these topics is that they are not newsworthy and you cannot play with it. It is so pervasively know that these are very demanding issue for education that nobody will be regarded as being “innovative” or “future oriented” in discussing these topics.

This leads me to my lesson learned from the current discussion on ChatGPT: We can see the appearance and public but also scientific reaction to it as a special case of agenda setting or agenda melding. While agenda setting theory (McCombs, Shaw, & Weaver, 2014) focuses on how (mass) media select topics and transmit them to the public with political or economical objectives in mind, agenda melding refers to the process by which members of the public chose and blend media agendas to fit their individual preferences. By firing the discourse on ChatGPT the educational community is actively pushing the agenda of companies with a commercial agenda and we should reflect if this is fitting to our individual preferences.

References

McCombs, M. E., Shaw, D. L., & Weaver, D. H. (2014). New directions in agenda-setting theory and research. Mass communication and society, 17(6), 781-802.

Sharples, M. (2022). Automated essay writing: an AIED opinion. International Journal of Artificial Intelligence in Education, 32(4), 1119-1126.

Sharples, M., & y Pérez, R. P. (2022). Story Machines: How Computers Have Become Creative Writers. Routledge.