Page 55 - AC-3-3
P. 55

Arts & Communication                                               AI in teacher education and critical thinking




                         A                                 B














            Figure 4. Example of an image created with AI. (A) Image created by a student (Leonarte, A., Master in Lifelong Research, University of Zaragoza, 2024).
            (B) Images created on based on picture in panel (A).
            Sources: Images provided by the author.

            hyper-realistic audiovisual content. This technology is   In addition, analysis of AI-created images requires
            mainly used to replace or manipulate faces and voices in   consideration of context and possible bias in their creation.
            videos so that they are perceived as authentic. The process   AI learns from large volumes of data, which often reflect
            involves training AI models with large amounts of visual   social biases or stereotypes. When examining these images,
            and auditory data of a specific person, which allows   it is necessary to reflect on the ideologies or values that may
            algorithms to learn their facial expressions, gestures, and   be implicit in them, as well as the possible omissions or
            voice. With this information, deepfakes can overlay these   exaggerations they present. This type of analysis deepens
            features on other images or videos, generating a convincing   critical understanding of how technology can replicate or
            imitation. The deepfake concept was designed in 2017 by a   amplify certain narratives and strengthens the ability to
            user of the Reddit website. This user founded a forum titled   question the biases present in algorithmic systems.
            deepfake to establish a discussion community about the   On the more negative side, there is the notion of false
            use of AI programs for the creation of deepfakes in the field   documentation. At present, there are websites that, upon
            of pornography. Even though deepfakes are mainly applied   payment of a certain amount for each processed image,
            in this area, deepfakes have also started to be virtualized   undress whoever the users request in a defamatory manner.
            for political purposes. 26                         What is striking is that these apps were not configured for
              The goal  of deepfakes is usually to mimic reality so   the treatment of male subjects, as the website in question
            accurately that they deceive the viewer. While they have   itself pointed out in its beginnings. Hence, the violence is
            legitimate applications, such as in films to recreate actors   not only in the aggression toward a body but focuses only
                                                                              27
            or in artistic projects, they are better known for their   on the female body.  At the same time, there is the creation
            malicious use. These uses include the creation of fake   of pedophilic content, also punishable by law. There are
            news or disinformation, falsification of public statements   already  several  municipalities  in  Spain  where  cases  of
            by important figures, fraud, and non-consensual    deepfake pornography have been reported.  The case of
            pornographic content. The ability of deepfakes to deceive   Almendralejo (Badajoz) was the first to appear, in which
            lies in their high level of realism, specifically designed to be   minor victims were stripped naked with an AI application,
            indistinguishable from reality.                    with its subsequent dissemination in different groups of
                                                               both WhatsApp and Telegram. 28
              In  addition,  analysis of  AI-created  images requires
            consideration of context and possible bias in their creation.   On the adult side, this type of representation is also
            AI may reflect cultural, racial, or gender biases present   found  in pornographic  key,  in  which  world  leaders  are
            in the data it was trained on, perpetuating stereotypes or   also defamed through this strategy, where their identity
            generating problematic content. When examining these   is impersonated and they appear in indecent attitudes, as
            images, it is necessary to reflect on the ideologies or values   happens with famous actresses and popular people of the
            that may be implicit in them, as well as the possible omissions   culture (Figure 5).
            or exaggerations they present. This type of analysis deepens   We found AI-created models, through impersonation,
            the critical understanding of how technology can replicate   that seduce viewers and generate economy by selling
            or amplify certain narratives and strengthens the ability to   products demanded by the target audience. We also found
            question the biases present in algorithmic systems.  cases where AI intervenes in people’s personal and intimate


            Volume 3 Issue 3 (2025)                         6                                doi: 10.36922/ac.5047
   50   51   52   53   54   55   56   57   58   59   60