What are the implications of 'visually orientated algorithms' for photojournalism?

近年来，记者在新闻编辑室里加入了一系列旨在使他们的工作自动化的软件。许多自动化系统执行的任务涉及大量结构化数据，例如体育和财经新闻，它们往往忽视了涉及更具创造性分析的领域。随着机器学习和自然语言生成技术的进步，这种情况开始发生变化。随着这些技术的应用越来越广泛，人们已经做出了一些努力来理解这些技术对书面新闻的影响。相比之下，这些技术对视觉新闻的影响，特别是对摄影新闻的影响，作为一个话题完全被忽视了。然而，面向视觉的算法对视觉新闻的变革潜力，就像面向文本的算法对书面新闻的变革一样。对这一领域缺乏研究可能反映了这样一个事实，即支持视觉导向算法的技术仍然落后于那些专注于分析更多定量数据和文本的技术，但人工智能和机器学习的最新进展意味着这一差距正在迅速缩小。认识到围绕这些主题的实证研究存在差距，我希望调查这些视觉导向算法对新闻摄影的当前和潜在未来影响，我将其描述为包括业余和专业制作的摄影和视频，用于说明或解释新闻和时事。此外，我想回顾一下这些影响是积极的还是消极的，并考虑可以采取哪些措施或步骤来预测或减轻一些消极后果。我将特别考虑以下几个核心问题。这些技术对于社会新闻摄影的目的意味着什么？摄影新闻的角色传统上被理解为通过单一的视觉人工制品报道新闻，“在公共文化中被每个人广泛认可”。我将质疑这些技术是否会导致这种转变，也许会朝着更加分散和个性化的方向发展。在前一个问题的基础上，我将问这些技术对公众对新闻摄影的信心意味着什么。该领域的声誉已经被广泛使用的错误归属图像、图像处理软件以及少数摄影师和编辑的道德不端行为所打击。更大的自动化是否提供了更容易发现和揭露这些滥用的承诺，或者它可能会无意中加剧它们？我将考虑这些技术对摄影新闻经济的影响。我想问的是，这些系统是否会提高新闻工作的效率，将记者从乏味而耗时的工作中解放出来，或者它们是否会威胁到技术性失业。这将包括考虑这些技术如何加剧已经被视为破坏新闻业经济模式的趋势。我将研究这些技术如何加剧视觉新闻中已经存在的偏见和结构性不平等。例如，考虑到该领域有问题的表示历史，可能有偏见的算法训练数据会进一步加剧这些趋势。我还会考虑到开发者有意融入偏见的危险，这些偏见旨在进一步推动媒体所有者的目标，并利用日益两极分化的政治和媒体环境。

In recent years, journalists have been joined in newsrooms by an array of software designed to automate aspects of their work. Many of these automated systems perform on tasks involving large quantities of structured data, for example sports and financial journalism, and they have tended to neglect areas involving more creative analysis. This situation is starting to change with advances in machine learning and natural language generation technologies. Some efforts have been made to understand the implications of these technologies for written journalism as their adoption becomes more widespread. By contrast the implications of these technologies for visual journalism, and photojournalism specifically, remain entirely neglected as a topic. Yet visually orientated algorithms have the potential to be every bit as transformative for visual journalism as their text orientated counterparts are now proving themselves to be for written journalism. This lack of inquiry into this area likely reflects the fact that the technology that underpins visually orientated algorithms still lags several steps behind those focused on analysing more quantitative data and text, but recent advances in artificial intelligence and machine learning mean that this gap is now rapidly closing. Recognising a gap in the empirical research around these topics, I wish to investigate the present and potential future implications of these visually orientated algorithms for photojournalism, a field which I characterise as encompassing both amateur and professionally produced photography and video used to illustrate or explain news and current events. Further to this I want to review whether these implications are broadly positive or negative, and to consider what measures or steps might be put in place to anticipate or mitigate some of the negative consequences. In particular, I will consider the following core questions. What do these technologies mean for the purpose of photojournalism in society? Photojournalism's role has traditionally been understood as one of reporting the news through singular visual artefacts broadly 'recognised by everyone within a public culture' I will question whether these technologies cause a shift away from this, and perhaps towards something far more fragmented and individualized. Building on the previous question, I will ask what these technologies mean for public confidence in photojournalism. The field's reputation has already been battered by the widespread use of misattributed imagery, image manipulation software, and ethical misconduct by a minority of photographers and editors. Does greater automation offer the promise of more easily catching out and exposing these abuses, or might it inadvertently exacerbate them? I will consider what the implications of these technologies are for the photojournalistic economy. I would ask whether these systems offer to make it more efficient, liberating journalists from banal and time-consuming tasks, or whether they might threaten technological unemployment. This would include considering how these technologies could aggravate trends already seen as undermining journalism's economic model. I would examine how these technologies might exacerbate biases and structural inequalities that already exist in visual journalism. For example, considering the problematic history of representation in the field, might biased algorithm training data exacerbate these tendencies further. I would also consider the danger of developers incorporating intentional biases designed to further the aims of media owners and to exploit an increasingly polarised political and media landscape.