One novel example and/or perspective in support of, “Why the learning account fails,” that the lack of viable ideographies is a learnability issue, is the impressive ability of humans to recognize and memorize facial features and accurately and reliably connect those to related identities. Faces are an aggregation of graphical features. Individuals are capable of learning and remembering hundreds of faces and facial expressions over decades and can identify extremely minor differences in facial features starting from a very young age (Nelson, Morse, & Leavitt, Reference Nelson, Morse and Leavitt1979). Human facial recognition is robust additional evidence that we are fully capable of processing and storing graphical features as well as phonemes. The challenge or failure of sophisticated artificial intelligence algorithms in correctly identifying faces compared to humans is a prime example of human ability in the graphical realm (Cavazos, Phillips, Castillo, & O'Toole, Reference Cavazos, Phillips, Castillo and O'Toole2021; Phillips et al., Reference Phillips, Yates, Hu, Hahn, Noyes, Jackson and O'Toole2018). This is not only an ability that humans are skilled at learning, it is also a highly evolved inherent skill crucial to maternal–infant bonding and successful early development. Brain regions critical for the identification of faces exhibit increased activity in new mothers compared to nulliparous women and these changes are closely linked to empathetic concern (Rigo et al., Reference Rigo, Kim, Esposito, Putnick, Venuti and Bornstein2019; Zhang et al., Reference Zhang, Rigo, Su, Wang, Chen, Esposito and Du2020). It is possible that the failure of some ideographies is because of a lack of emotional relevance, where an increased focus on emotional impact in the development of graphic codes will result in enhanced adoption.
Although most cognitive traits possess modest heritability, twin studies of facial recognition reveal a high-genetic contribution (Wilmer et al., Reference Wilmer, Germine, Chabris, Chatterjee, Williams, Loken and Duchaine2010). However, this ability does not strongly correlate with measures of visual and verbal recognition, leading to the conclusion that facial recognition is both specific and highly heritable, underscoring its importance to our survival as a mammalian species highly dependent on parental care and social interaction. Although this specificity of facial recognition could be seen as support for learnability-based theories of challenges to ideographies, it can also be interpreted as a lack of relevant and valid tools to reliably assess general visual recognition (Diaz-Orueta, Rogers, Blanco-Campal, & Burke, Reference Diaz-Orueta, Rogers, Blanco-Campal and Burke2022). Along these lines, perhaps what is needed for the development of effective ideographies is a focus on less abstract, more organic shapes and features commonly found in human faces. Given the findings from studies of mothers and the heritability of facial recognition, it could indeed be that emojis are the future of effective ideography.
Neuroimaging analysis presents an example in support of the crucial role of standardization in the lack of adoption of ideography. Functional neuroimaging involves the processing and interpretation of brain activity that is often depicted solely by graphical representation. Similar to facial recognition algorithms, neuroimaging analysis often involves numerous assumptions and compromises and the never-ending methods development in this field has led to overall inconsistency (Poldrack et al., Reference Poldrack, Baker, Durnez, Gorgolewski, Matthews, Munafò and Yarkoni2017). The lack of standardization in the analysis of neuroimaging data is widely acknowledged as a key reason for poor replication and repeatability in neuroimaging studies in the past (Kennedy et al., Reference Kennedy, Abraham, Bates, Crowley, Ghosh, Gillespie and Travers2019), similar to the weak agreement on the meaning of various emojis. Accordingly, recent efforts to broadly increase standardization in neuroimaging (Niso et al., Reference Niso, Botvinik-Nezer, Appelhoff, De La Vega, Esteban, Etzel and Rieger2022), functionally increasing agreement on the meaning of graphical images of brain activity, may substantially enhance replication across the field. At the individual level, the standardization in functional neuroimaging that is focused on changes in brain activity has accelerated the learning process within and across trainees, resulting in the increased adoption of what could be referred to as a very complex form of ideography.
An integral aspect of neuroimaging analysis is the discrimination of patterns of functional brain activity versus patterns indicative of statistical noise/variability because of various confounds such as electrical interference or subjects moving during scans. This can occur through two methods, one based on organic human learning and training, the other based on a form of artificial intelligence, deep learning. The identification and extraction of these confound-based patterns can be accomplished through the training of individuals and/or the use of an image recognition program. Similar to facial recognition, the best performance is often obtained through the use of both image recognition programs and subsequent human confirmatory inspection. However, variability in training and human performance has adverse effects on reproducibility and current efforts are focused on the application of deep learning to neuroimaging analysis (Abrol et al., Reference Abrol, Fu, Salman, Silva, Du, Plis and Calhoun2021). Perhaps we should consult ChatGPT on how best to design a self-sufficient graphic code?
You have
Access
One novel example and/or perspective in support of, “Why the learning account fails,” that the lack of viable ideographies is a learnability issue, is the impressive ability of humans to recognize and memorize facial features and accurately and reliably connect those to related identities. Faces are an aggregation of graphical features. Individuals are capable of learning and remembering hundreds of faces and facial expressions over decades and can identify extremely minor differences in facial features starting from a very young age (Nelson, Morse, & Leavitt, Reference Nelson, Morse and Leavitt1979). Human facial recognition is robust additional evidence that we are fully capable of processing and storing graphical features as well as phonemes. The challenge or failure of sophisticated artificial intelligence algorithms in correctly identifying faces compared to humans is a prime example of human ability in the graphical realm (Cavazos, Phillips, Castillo, & O'Toole, Reference Cavazos, Phillips, Castillo and O'Toole2021; Phillips et al., Reference Phillips, Yates, Hu, Hahn, Noyes, Jackson and O'Toole2018). This is not only an ability that humans are skilled at learning, it is also a highly evolved inherent skill crucial to maternal–infant bonding and successful early development. Brain regions critical for the identification of faces exhibit increased activity in new mothers compared to nulliparous women and these changes are closely linked to empathetic concern (Rigo et al., Reference Rigo, Kim, Esposito, Putnick, Venuti and Bornstein2019; Zhang et al., Reference Zhang, Rigo, Su, Wang, Chen, Esposito and Du2020). It is possible that the failure of some ideographies is because of a lack of emotional relevance, where an increased focus on emotional impact in the development of graphic codes will result in enhanced adoption.
Although most cognitive traits possess modest heritability, twin studies of facial recognition reveal a high-genetic contribution (Wilmer et al., Reference Wilmer, Germine, Chabris, Chatterjee, Williams, Loken and Duchaine2010). However, this ability does not strongly correlate with measures of visual and verbal recognition, leading to the conclusion that facial recognition is both specific and highly heritable, underscoring its importance to our survival as a mammalian species highly dependent on parental care and social interaction. Although this specificity of facial recognition could be seen as support for learnability-based theories of challenges to ideographies, it can also be interpreted as a lack of relevant and valid tools to reliably assess general visual recognition (Diaz-Orueta, Rogers, Blanco-Campal, & Burke, Reference Diaz-Orueta, Rogers, Blanco-Campal and Burke2022). Along these lines, perhaps what is needed for the development of effective ideographies is a focus on less abstract, more organic shapes and features commonly found in human faces. Given the findings from studies of mothers and the heritability of facial recognition, it could indeed be that emojis are the future of effective ideography.
Neuroimaging analysis presents an example in support of the crucial role of standardization in the lack of adoption of ideography. Functional neuroimaging involves the processing and interpretation of brain activity that is often depicted solely by graphical representation. Similar to facial recognition algorithms, neuroimaging analysis often involves numerous assumptions and compromises and the never-ending methods development in this field has led to overall inconsistency (Poldrack et al., Reference Poldrack, Baker, Durnez, Gorgolewski, Matthews, Munafò and Yarkoni2017). The lack of standardization in the analysis of neuroimaging data is widely acknowledged as a key reason for poor replication and repeatability in neuroimaging studies in the past (Kennedy et al., Reference Kennedy, Abraham, Bates, Crowley, Ghosh, Gillespie and Travers2019), similar to the weak agreement on the meaning of various emojis. Accordingly, recent efforts to broadly increase standardization in neuroimaging (Niso et al., Reference Niso, Botvinik-Nezer, Appelhoff, De La Vega, Esteban, Etzel and Rieger2022), functionally increasing agreement on the meaning of graphical images of brain activity, may substantially enhance replication across the field. At the individual level, the standardization in functional neuroimaging that is focused on changes in brain activity has accelerated the learning process within and across trainees, resulting in the increased adoption of what could be referred to as a very complex form of ideography.
An integral aspect of neuroimaging analysis is the discrimination of patterns of functional brain activity versus patterns indicative of statistical noise/variability because of various confounds such as electrical interference or subjects moving during scans. This can occur through two methods, one based on organic human learning and training, the other based on a form of artificial intelligence, deep learning. The identification and extraction of these confound-based patterns can be accomplished through the training of individuals and/or the use of an image recognition program. Similar to facial recognition, the best performance is often obtained through the use of both image recognition programs and subsequent human confirmatory inspection. However, variability in training and human performance has adverse effects on reproducibility and current efforts are focused on the application of deep learning to neuroimaging analysis (Abrol et al., Reference Abrol, Fu, Salman, Silva, Du, Plis and Calhoun2021). Perhaps we should consult ChatGPT on how best to design a self-sufficient graphic code?
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interest
None.