Page 16 - Can AI take education to a new level
P. 16

AI IN EDUCATION


     Home


     News                         “Building resilience, the ability to roll with the punches and man-  says that software-generated text can give itself away by being
                                age change at a fast pace – those are dispositions you can prac-      too consistent – it tends to lack human randomness. But detec-

     How the chip sector        tice and teach,” says Harrison.                                       tion tools can only generate a probability that an LLM was
     is gearing up for            School homework is an example of what educators called form-        used, and unlike with straight plagiarism, they cannot point to a
     the AI revolution
                                ative assessment, which  aims to monitor and reinforce learn-         likely source.
                                ing  during  a course. Using  LLMs in summative  learning,  such       “None of them can prove conclusively that the text was written
     How ISC2 aims
     to overcome                as exams and coursework at the end of a course, is particularly       by AI,” he says. Furthermore, as LLM tools are incorporated into
     cyber barriers             sensitive as it could allow students                                                              software such as Microsoft Word,
                                to  gain  a  qualification  or  a  higher                                                         their use may become as routine
     Editor’s comment           grade if they do so.                        there are potentIally three                           as spelling and grammar checking
                                  Jisc,  a  not-for-profit  agency  that                                                          is at present.

     Buyer’s guide              provides technology services across         ways for educators to deal                             Webb  sees three ways for edu-
     to the future of           UK further and higher education, is                                                               cators to deal with  AI in assess-
     business software                                                   wIth aI In assessment: avoId It,
                                researching how LLMs are affecting                                                                ment: avoid it, outrun it or embrace
                                different types of assessment. It has                                                             it.  Avoidance  means using highly
     Harnessing large                                                           outrun It or embrace It
     language models            already found that  multiple choice                                                               controlled  environments  such as
     for education              quizzes are particularly vulnerable.                                                              exams, but these don’t work for all
                                  Michael Webb,  director of  technology and analytics  at Jisc,      assessment. Trying to outrun AI by using formats that services
                                says that previously, someone trying to cheat on an online quiz at    can’t  handle  is  likely  to  be  a  short-term  fix  given  how  fast  the
                                least had to use a search engine, potentially learning something      technology is developing, with ChatGPT’s supplier OpenAI add-
                                in the process, but LLMs short-circuit even that. “Anything that      ing support for diagrams in March.
                                has been done before, ChatGPT is really good at,” he says.             He believes educators will usually be better off embracing
                                                                                                      LLMs. This can include setting guidance and rules for their use,
                                Ai detection cApAbility                                               something  many  universities  have  already  done,  and  requir-
                                Many institutions use plagiarism detection tools such as TurnItIn,    ing students to declare they have used them – although if AI
                                run by a Californian-based company of the same name, which            becomes ubiquitous in word processing, this could be a tempo-
                                aims to identify material that has appeared previously. In March      rary measure. Assessments could be based on new data from
                                2023, the company launched an AI detection capability. Webb           ongoing or yet-to-be-published academic research, or students




                                                                                                                                    computerweekly.com 21-27 November 2023 16
   11   12   13   14   15   16   17   18   19