Twitter, Human Technology FoundationLinkedIn, Human Technology Foundation
Working group on the impact of generative AI systems

The context

Since the deployment of ChatGPT, the OpenAI's intelligent conversational agent, generative AI systems (Gen AIS) have continued to occupy discussions and be the subject of a massive production of work. However, it has to be said that these works tend to be of widely varying quality and relevance, demonstrating that although the subject has been widely democratised, it remains often poorly understood.

The reason why Gen AIS have attracted so much attention is precisely because they unquestionably offer opportunities that organizations, both public and private, can seize in order to optimize their activities and remain competitive.

This enthusiasm is also the result of concerns linked to the development of these systems, which inevitably raises ethical questions. By generating content that is disturbingly realistic, these systems help to blur the distinction between reality and artificiality, impacting on perceptions but also raising questions about their malicious use by certain players.

The context of rapid technological change and regulatory uncertainty also encourages ethical questioning. These questions may be generally related to the governance of these systems and their impact on human beings, but they are above all very specific questions about trust in the content produced by AIS; their deployment and management within organizations; and the impact, both negative and positive, of these technologies on organizations and their employees.

It is this last point that we felt should be addressed, and which has led to the establishment of a cross-sector working group (WG) on the impact of Gen AIS.

The Gen AIS  WG's goal

In the current context, it is essential to rigorously identify the positive and negative impacts of Gen AIS in order to develop best practices to capitalise on the former and limit, if not avoid, the latter.

To help decision-makers meet these challenges, the Human Technology Foundation (HTF) has launched a French-speaking working group bringing together international players (developers, integrators and users of Gen AIS). The aim of this WG is to draw up as exhaustive a map as possible of the positive and negative impacts of Gen AIS, to organize them using a practical taxonomy, and finally to draw up best practices for adoption.

The AIS Gen WG now brings together 38 public and private organizations represented by 76 contributors from 3 continents. The diversity of these points of view enables us to better define the scope of the impacts associated with Gen AIS.

The work of this working group, conducted under the rule of Chatham House, will be the subject of an analysis report intended exclusively for the contributors. A simplified version will be made public and presented to public policy decision-makers in the countries where the Foundation is present. An interim report is already available on our website.

This project has four strong points that make it unique. The project aims to be: 

  • Pragmatic: Best practices must be identified with a view to their evolution. They will therefore be updated in line with developments in technology, legal standards and societal expectations
  • Multi-stakeholder: the project brings together developers, integrators and users of Gen AIS, from a variety of public and private sectors
  • Apartisan: the initiative is apartisan to ensure that solutions are found that are suitable for all participants
  • Responsible: the project is based on responsibility and practical ethics

By focusing as closely as possible on the needs of the players involved and adapting to technological developments, the work carried out by the WG will fuel collective thinking and facilitate the responsible development of Gen AIS.

The project is being carried out in two phases, with tangible benefits for the players involved. 

The first phase ran from January to June 2024 and consisted of hearings and monthly plenary meetings supported by chamber work. The main outcome of this phase was the mapping of more than 160 use cases for generative AIS, on the basis of which a taxonomy of positive and negative impacts was drawn up, followed by a prioritization of these impacts. Based on this work, best practices and recommendations have been defined to help contributors to the Gen AIS WG to more easily identify the relevant impacts in order to capitalize on the positive impacts while limiting the negative impacts. 

The second phase of the project will begin in June 2024, with the establishment of a permanent WG to monitor technologies and standards. The aim will be to perpetuate the work carried out during the first phase and ensure quarterly updates of use cases, impacts and best practices. The advantage of such an approach lies in the ability it gives contributors to adapt as closely as possible to developments in terms of technology, legal constraints and societal expectations.  

The deliverables

So far, the WG has produced three key deliverables:

  1. Mapping of use cases 

The mapping of Gen AIS use cases was drawn up through hearings with WG participants and through research work leading to the identification of 167 relevant use cases. This mapping has been enriched and updated on an ongoing basis, benefiting, among other things, from feedback from contributors during plenary meetings.

  1. Taxonomy of impacts

Based on the mapping of use cases, a list of positive impacts and a list of negative impacts were identified. These two lists have been organized into an easily exploitable taxonomy. A multiscalar approach (individual, organizational, societal and global levels) was adopted to make the taxonomy easier to use. 

  1. Ranking positive and negative impacts 

As we expanded the taxonomy during our plenary sessions, we noticed that the impact categories identified applied to Gen AIS as well as to most existing digital technologies. By focusing on the specific features of Gen AIS, we have prioritized the impacts most exacerbated by these new technologies. 

Three types of impact were of particular interest: on the environment, on skills, and on productivity. The creation of a working group focusing on environmental impacts was approved at plenary session 5 on 16 May 2024.

At the end of this process, we identified five major peripheral findings: 

  1. In terms of impact, it is difficult to make a strict distinction between generative AIS and traditional AIS. 
  2. The differences between traditional AIS and Gen AIS seem to be based on two elements: their ability to blur the boundary between the real and the artificial, and the increase in quantity and quality compared with traditional AIS. 
  3. Ethical thinking is still heavily structured by compliance, which limits its understanding, application and scope. 
  4. The use cases for Gen AIS remain limited in organizations compared with the possible use cases identified. 
  5. There are significant differences in approach between geographical areas, and therefore between cultural areas, as regards both the use of Gen AIS and the assessment of their impact, particularly in terms of ethical acceptability and desirability. 

As we approach the end of this first phase of our work, we can see that Gen AIS are presenting a growing number of use cases whose impacts are as varied as those of other digital technologies.

The deliverables will be living objects that will need to be updated on a regular basis in order to benefit from adaptable and relevant best practices. The work of the WG will also continue with the aim of identifying weak signals concerning technical and normative developments surrounding Gen AIS, in order to move away from a reactive approach towards one of anticipation.

The work of the WG will be completed by the reflections of the WG focused on environmental impacts, as well as by specific research work supported by expert consultations. The impacts on skills and productivity will also be the subject of discussions that will be shared through technical notes.

The project as a whole will enable contributors to benefit from up-to-date information, shared experience, good practice and potentially early warnings that will facilitate decision-making on the adoption and use of Gen AIS, making decisions more relevant and therefore more effective.

The full report is available on our website : https://www.human-technology-foundation.org/news/report---generative-ai-systems

Related Articles