Page 108 - FerMUN 2020 - Futurecasters Global Young Visionaries Summit, 8th-10th January 2020
P. 108

ii.   Ensuring the smooth running and proper functioning of the programs;

                   2.  Affirms  that  behavioural  analysis  software  could  lead  to  misinterpretations  of  facial

                      signals, hence the need for validation of results by humans as a second-step safety
                      measure, meaning that effective system monitoring mechanisms should be put in place
                      to avoid unintentional mistakes;


                   3.  Strongly  encourages  the  establishment  of  a  committee  under  the  direction  of  the
                      United Nations called the United Nations Committee for Ethical Artificial Intelligence

                      (UNCEAI), which will perform tasks including
                          a.  Verifying the state of AI deployment in education systems in all States,
                          b.  Informing  the  ministry  of  education  of  any  relevant  State  in  case  of  non-
                              compliance with the present resolution;


                   4.  Supports the establishment of a system within the UNCEAI which will be responsible for:
                          a.  Maintaining equal decision-making power for all member States,

                          b.  Rewarding States whose policies demonstrate responsibility and awareness of
                              the  risks  and  potential  of  AI  with  official  recognition,  thus  granting  them  an
                              important role in the development of new guidelines;


                   5.  Reaffirms that the datasets fed into behavioural analysis software used for educational
                      purposes should be directly available to governments to ensure the safety and security
                      of students in the country;


                   6.  Invites member States to reduce the bias in output of AI systems in order to achieve
                      unbiased and unquestionable outcomes from the ethical perspective by:

                          a.  Processing the datasets beforehand in such a way that every decision made is
                              equal to every group, including ensuring that:
                                i.   The datasets used are varied,

                                ii.   Datasets with similar attributes lead to similar results,
                          b.  Incorporating  the  principle  of  fairness  into  the  training  process  itself,  called
                              “counterfactual fairness”, which assigns different attributes to individuals in a
                              counterfactual world with the aim of comparing the results in the actual world;


                   7.  Suggests  collaboration  between  AI  and  humans  in  order  to  reduce  mismatches,
                      introducing individuals who will:

                          a.  Run algorithms,
                          b.  Compare results,





                                                                                                      107
   103   104   105   106   107   108   109   110   111   112   113