Introduction
Image and Gaze

The aesthetic emotion, which gives rise to the impression of beauty, is undoubtedly universal in humans.

Edgar MORIN (2013, p. 13)

The proliferation of images in the modern world increasingly forces us to select photos from among a vast set. The task of choosing images to illustrate an article, a book cover, a poster, or one that represented an event or a voyage used to be entrusted only to professionals in the fields of publishing, communication, archives, or professional photographers and collectors. Today, it is incumbent on each of us to decide what we wish to preserve and what to forget, what we wish to send, post on the Internet, delete forever, or to archive (in the improbable event of us ever revisiting it). Managing photo archives is a real bane for many us1. The step of selecting images has always been assumed to be delicate and of great importance. In the professional context, it is entrusted to those who are renowned as experts or in a position of authority.

Let us pause for a moment to examine the mechanisms that are activated when we make decisions about sorting and selecting images. Where does the beauty of the image fall in our hierarchy?

It would seem that the primary mechanisms that are activated when selecting photographs can be divided into three broad categories:

- why the document is of interest, that is, its ability to draw and hold our attention by relating the document to contexts familiar to us;

- the surprise factor, that is, contrary to the previous point, its ability to give us a novel visual or cognitive experience by bringing in an unexpected contribution;

- beauty, that is, the pleasure it brings us, independent of its content, through the arrangement of its elements.

The last point is what we will be looking at, exclusively, in this book.

In this framework, the first two points are the result of associative impressions, as used by G.T. Fechner, who used this distinction from the 19th century (Fechner 1871). It often happens that the same image associates several of these registers, with its attractiveness being heightened, but with the contribution of each register of attention being less clear. However, with respect to our decision-making, the contributions of interest, surprise and beauty would seem to evolve2, in independent or orthogonal spaces, as used by Gärdenfors (2000), that is, without any intimate influence on each other. They are therefore evaluated separately by our consciousness, and then probably combined into a single score that ultimately makes us prefer one image to another in a heuristic choice that is difficult to express, but which is likely to follow the empirical decision-making schemas proposed by Tversky and Kahneman (1981).

Before we move away from them, let us briefly study the first two domains. In the field of image processing, these are sometimes referred to by other names as well: interestingness3, memorability4, unusualness5, or popularity6; we will illustrate these with examples (Isola et al. 2011; Gygli et al. 2013; Amengual et al. 2015).

Mechanisms related to interest

These mechanisms have been of particular interest in studies among experimental psychologists from the 1920s onwards. The work produced by these psychologists is often brought together under the umbrella term relevance theory (Sperber and Wilson 2004).

Interest or relevance, that is, a complex psycho-physiological experience, may be produced by:

- external stimuli (environmental stimuli transmitted by our senses) or internal stimuli (biochemical, created by our cortex). It is therefore a passive process;

- or through reasoning, that is, cognitive processes that produce new elements of knowledge from a particular context. This is then an active process.

W. James' work (Lange and James 1922) pioneered studies on the relation between emotion and how a person appraises a situation. The importance of arousal and the order of the various steps involved (arousal, emotion, appraisal) has also been well documented. This research has served as a guide for various studies conducted 50 years later in the field of aesthetics.

However, it was Sperber and Wilson (2004), above all, who made it possible to construct a relevance theory, bringing together all the cognitive baggage of the recipient and not only their linguistic knowledge, as suggested by W. James. In this context, Dessalles (2008) suggested a more quantitative measure of relevance and endowed it with predictive capacities using an original mathematical model.

If we want to apply this relevance theory to the interest aroused by a collection of images, it is useful to define two extreme cases that may be approached in different ways:

- universal interest refers to themes that are often displayed in society and media and transmitted through context and culture: such-and-such an actor or sportsperson, a car, monument or an event that is explicitly and regularly covered. The popularity of these themes can today be measured using mediametry tools (Hsieh et al. 2014; Fu et al. 2014): frequency of exposure on television, the number of instances on the Internet, "clicks" or "likes" on social media, etc. (Figure I.1(a));

- personal interest brings one closer to the images because the themes are deeply connected to individual life, more specifically, the viewer's personal life: "my" family, "my" city, "my" work, etc. The evaluation of relevance then takes the conventional forms proposed in Dessalles (2013), for example, which involves variables of space and time that decrease more or less rapidly, and relations in the degree of proximity (e.g. in a family tree or a company flow chart) (Figure I.1(b)).

Figure I.1 An image's relevance may be related to the general context of knowledge within a community ((a) actors, sportspersons, objects that frequently appear in newspapers or social media), or, on the contrary, to the personal context of the observer ((b) members of the same family, vacation sites, hobbies, etc.).

Mechanisms associated with surprise

These mechanisms function a little differently from those mentioned above, as they become more effective when the document presented to us is more removed from the familiar (Figure I.2). Surprise or amazement (as this is a broader term, it may be more useful for us here as it covers a wider range of responses) can be narrowed down to various forms: humor, fear, perplexity, affection, disgust, etc.

It can be clearly seen that relevance, on the one hand, and surprise, on the other, are different from aesthetic qualities that motivate us to remember a photograph. It is also quite likely that they operate together to lead to a choice7, as can be seen in the two photos in Figure I.3. However, in the following sections, we will do our best to separate these two mechanisms. That is, we will strive to assess aesthetic qualities with relevance and surprise being equal. This will not be easy and it must be remembered that in many situations, the opinions shared during subjective tests are likely to have been somewhat confused on this point (Gygli et al. 2013). This is especially true when opinions are solicited from unknown and remote persons, as happens with evaluations on the Internet, for example.

Figure I.2 Amazement is another mechanism that leads us to pay particular attention to an image. We then strive to clearly identify the specific elements in the image that fall beyond the known schemas of our representation of the universe.

Pleasure versus arousal

The three motivations behind our attention when we study a photo, namely its beauty, our interest in the subject of the photo, and the surprise it may produce in us, are the result of a long, careful and well-reasoned observation. However, it is also possible to study the immediate effects on ourselves following a very brief viewing of a photo. We therefore seek to identify the most elementary biological effects, which involve only the most basic physiological activation and not the elaborate functioning of cognition and reasoning. This is done by presenting a photo to an observer for a fraction of a second and measuring a few physiological indicators that reflect two independent emotional reflexes: pleasure and arousal8. These experiments were first carried out by psychologists at the National Institute of Health (USA) and were then repeated by many other authors. This experiment was carried out by creating a database with a thousand reference photographs, the IAPS database (International Affective Picture System9) (Lang et al. 1999). They then conducted psychophysiological experiments that made it possible to place each image along a pleasure graph as a function of the arousal, revealing two clear orientations, which they named the direction of appetitive motivation and the direction of defensive motivation (Figure I.4).

Figure I.3 Both these photographs are indisputably remarkable for their aesthetic qualities. Photograph (a) (Portrait of Jean Cocteau by Irving Penn, 1948) holds our attention because we recognize a famous man ("universal interest" stimulus). Photograph (b) (Bruxelles by...